XI__Y_1_IIX__CI~II__-~~III~-l-
IIL-~W
GROUNDWATER FATE OF AROMATIC HYDROCARBONS
AT INDUSTRIAL SITES: A COAL TAR SITE CASE STUDY
by
Allison Ann MacKay
S.M., Civil and Environmental Engineering, Massachusetts Institute of Technology, 1993
B.A.Sc., Engineering Science, University of Toronto, 1991
Submitted to the Department of Civil and Environmental Engineering
In Partial Fulfillment of the Requirements of the Degree of
DOCTOR OF PHILOSOPHY
in Civil and Environmental Engineering
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
February 1998
© Massachusetts Institute of Technology. All rights reserved.
Signature of the Author
Department of Civild
Certified by
__
Environmental Engineering
January 16, 1998
\i
Philip M. Gschwend
Professor of Civil and Environmental Engineering
Thesis Supervisor
Accepted by
Joseph M. Sussman
Graduate Studies
on
Committee
Chairman, Departmental
Flzi3C1I
ir~i---.
~-l---.-~
r- ruluxrri~---~
-~---r^r*--rxr~c.
-I-~~I*-L---~LI- I--^-^X-II-IP~
II___I__I__I .- -.-1~-- -~--~~~~~--~11~1~
GROUNDWATER TRANSPORT OF AROMATIC HYDROCARBONS
AT INDUSTRIAL SITES: A COAL TAR CASE STUDY
By
Allison Ann MacKay
Submitted to the Department of Civil and Environmental Engineering on
January 16, 1998 in Partial Fulfillment of the Degree of
Doctor of Philosophy in Civil and Environmental Engineering
Abstract
The fate of groundwater contaminants in anthropogenic fill materials was investigated
at a coal tar site, Site YYZ. This site was representative of other contaminated sites with a
history of industrialization at which wastes from process operations form the local subsurface
solids. The solids composing the groundwater-bearing unit at Site YYZ were reactants (oil),
byproducts (tar, coke) and wastes (gas purification box waste) used and produced during 100
years of gas manufacture operations at this site. The in situ groundwater transfer and reaction
processes acting upon aromatic hydrocarbons in the subsurface at this site were hypothesized
by comparing the groundwater fingerprints of individual compound concentrations to
measured aqueous concentrations of these hydrocarbons in coal tar-equilibrated water. In
general, the groundwater concentrations agreed with tar-equilibrated aqueous concentrations,
indicating the source of aromatic hydrocarbons in the groundwater was equilibrium dissolution
of the nonaqueous phase liquid tar.
The first field evidence of colloid-enhanced solubilization of hydrophobic organic
colloids was found at Site YYZ. At some monitoring wells, groundwater polycyclic aromatic
hydrocarbon (PAH) concentrations were greater than measured for aqueous equilibrium with
tar by a factor which increased with compound hydrophobicity. Two thirds of the PAH mass
in excess of dissolved solubility was associated with particles that could be settled from
solution over 5 months. The remaining excess PAH mass was associated with 4 mgc/L
suspended organic carbon that was stable over 5 months, but could be precipitated at pH1,
suggesting that PAHs were associated with humic acid-like molecules. About 5 mgc/L of
humic and fulvic acids were present in the groundwater at all monitoring wells sampled. The
presence of colloids in the groundwater will increase the off-site flux of hydrophobic PAHs
over flux estimates assuming only dissolved equilibrium with coal tar.
Evidence of bioattenuated xylene, naphthalene and methylnaphthalene concentrations
was found at the shallow monitoring wells. The compound depletion patterns and
groundwater ion concentrations were consistent with aromatic hydrocarbon removal by
sulphate reducers. Biodegradation acted to decrease the off-site flux of these compounds,
relative to tar-water equilibrium at Site YYZ.
Solid-water partitioning to carbon-containing anthropogenic fill solids isolated from
Site YYZ was also quantified. The overall partition coefficients for fill solids mixtures were
described by summing the sorption contributions of the individual materials, using sorbentspecific partition coefficients. Predictions of overall partitioning made on a carbon basis,
assuming a natural organic carbon partition coefficient, were up to two orders of magnitude
different from measured values. An octanol water partition coefficient (Kow)-based linear free
energy relationship for predicting sorbent-specific partitioning to wood was developed from an
investigation of monoaromatic hydrocarbon sorption to wood chips, where log Kgnm =
0.71 log Kow + 0.08. Accurate predictions of groundwater transport through anthropogenic fill
solids at industrial sites must account for the composition of the fill matrix.
Thesis Supervisor: Dr. Philip M. Gschwend
Title: Professor
^I~
I~------- I--~-~IC---L~;P.-~-----L--~I~.^
~I_
-_~ ---^~~
_.~_ly_
_~~^--
t11111 .)I^ LI^Y LIII~-Y--L---.I
- L
~
III~
I*CI-~~-I
II__
__^I1I~_1I
---*~
----~- -1TI I1~I_~~~_
~~_~ ~sU
Acknowledgements
Funding for this research was provided by Baltimore Gas and Electric Company.
This thesis could not have been completed without the help of many others.
Thanks to Herb Hoffman, formerly of Baltimore Gas and Electric, Ian MacFarlane of EA
Engineering, Science and Technology, and Rick Walden and Lee Malinowski of Baltimore
Gas and Electric for technical and field support. Peter Dow, Mike and Ed of Environmental
Drilling Inc. applied their expertise to obtain cores and install wells in the uncohesive fill.
Extra big thanks to John MacFarlane, Freddi Eisenberg and Chris Swartz for their help and
companionship while field sampling in the unseasonable, inclement weather I seemed to
attract.
Thanks to my thesis advisor Phil Gschwend for helping me to strive for excellence in
scientific questioning and for clarity in my thought expression. Phil and the other members
of my thesis committee, Harry Hemond and Dennis McLaughlin provided many insights to
help bring stacks of raw data together into a coherent story.
Thanks to members of the Gschwend group, Parsons Lab and the Gas Turbine Lab for some
necessary diversions from research during my years at MIT, and for providing, perhaps
unknowingly, sparks of motivation along the way. In my final months, I enjoyed the
camaraderie of Tory Herman, Randi Carlson, and Chris Swartz who also suffered from the
common task of having to manipulate the English language, instead of lab equipment.
My development as a scientist benefitted greatly from interesting conversations with Tory
Herman and Lynn Roberson about "the way everything else works"
I especially thank my best friend Ken Gordon for his unwavering support, awesome cooking
and the coolest camping trips ever. What a challenge we have met together - double PhDs!!
~~~~---II------~- ~-~""-Y~~~' ~ ~~-~""-~"~~"I~V~
1~'~1"1~-~ -I~~--1
Discussion
Changes in Groundwater Chemistry from an Induced
Groundwater Gradient
Fate of Coal Tar Constituents at Site YYZ
Equilibrium Coal Tar Dissolution
Facilitated Transport
Biodegradation
Summary of Results and Implications for Off-Site Transport and
Remediation
References
Chapter 3
MECHANISMS OF GROUNDWATER SOLUBILITY ENHANCEMENTS OF
AROMATIC HYDROCARBONS AT A COAL TAR SITE
Abstract
Introduction
Methods
Chemicals
Sample Treatments
Fractionated Extractions of Groundwater
Fluorescence Quenching
Organic Carbon Measurements
Removal of Organic Carbon
Calculation of Partition Coefficients
Results and Discussion
Fractionated Extractions of Groundwater
Fluorescence Quenching
Correlation of Organic Carbon with Enhancement Factors
Conclusion
References
60
60
64
64
66
76
81
83
87
88
89
91
91
91
91
92
93
93
94
94
95
97
104
107
108
Chapter 4
HYDRAULIC PROPERTIES OF FILL SOLIDS
Abstract
Introduction
General Characteristics of Fill Solids
Site YYZ Hydraulics
Conceptual Picture of Site YYZ Hydrology
Hydraulic Conductivity
Groundwater Velocity
References
110
111
112
112
114
114
117
120
126
Table of Contents
Abstract
3
Acknowledgements
5
Table of Contents
7
List of Tables
12
List of Figures
14
Chapter 1.
INTRODUCTION
Introduction
History of Manufactured Gas Production and the Nature of Site
Contamination
Research Outline
References
19
23
25
Chapter 2
AROMATIC HYDROCARBONS IN GROUNDWATER AT A COAL TAR
SITE
Abstract
Introduction
Site Description
Methods
Well Installation
Groundwater Sampling
Chemicals and Glassware
Volatile Compound Analysis
PAH Analysis
Verification of Compound Identities
Tar Analysis
Tar-Water Equilibration
Inorganic Compounds
Carbon Analysis
Surface Tension
Electron Microscopy
Results
Method Evaluation/Sample Quality
Groundwater Quality Parameters
Mineral Phases
Aromatic Hydrocarbons in Groundwater
26
27
28
31
33
33
34
36
37
37
39
39
39
40
40
40
41
41
41
46
49
51
17
18
~_I~----.L-(II1~IIIX
..
lX_ I_~---C
~1-1 --
~10
-1
-^- _._1_4~-^---~1---
Chapter 5
SORPTION OF HYDROPHOBIC COMPOUNDS TO FILL SOLIDS
Abstract
Introduction
Scope of investigation
Methods
Chemicals
Solids Collection
Solids Characterization
Fraction Organic Carbon
Nonaqueous Phase Liquids
Polycyclic Aromatic Hydrocarbons
Sorption Isotherms
Sorbents
Analysis
Fluorescence
Gas Chromatography
Tar Content
Elemental Analysis
Surface Area
Experimental
Mass Balance
Equations
Results and Discussion
Characterization of Anthropogenic Fill Solids
Sorption Isotherms
Natural Solids
Box Waste
Solvent-Extracted Box Waste
Coke Wastes
Sorbent Quantification
Conclusion
References
Chapter 6
SORPTION OF NONPOLAR ORGANIC COMPOUNDS TO WOOD
Abstract
Introduction
Wood Physiology
Chemical Composition
Physical Structure
Sorption of Nonpolar Organic Compounds to Wood and Wood
Components
now-U
127
128
129
130
131
131
131
132
132
132
132
132
132
133
133
133
134
134
134
134
135
135
138
138
141
143
145
147
151
153
157
160
164
165
166
167
167
169
171
Diffusion in Wood
Physically Hindered Diffusion
Homogeneous Retarded Diffusion
Scope of Investigation
Methods
Chemicals
Equilibrium Sorption Isotherms
Wood Sorption Kinetics
Equations
Sorption Isotherms
Sorption Kinetics
Results and Discussion
Equilibrium Sorption Isotherms
Wood-water partition coefficients
Linear Free Energy Relationship for Lignin-Water
Partition Coefficients
Kinetics of Wood Sorption
Experimental t1 /2 Values
Characteristic Diffusion Times - Phy sically Hindered
Diffusion
Characteristic Diffusion Times - Hoi mogeneous Retarded
Diffusion
Environmental Relevance
References
Chapter 7
SUMMARY OF RESULTS AND
FUTURE STUDY OF INDUSTRIAL SITES
Introduction
Summary of Results
Application of Results to Transport Calculations at Site YYZ
Areas of Further Investigation
General Approach to Remedial Investigations of Contaminated Sites
with a History of Industrial Activity
References
Appendix A
AQUEOUS SOLUBILITY OF AROMATIC HYDROCARBONS
IN EQUILIBRIUM WITH COAL TAR
Introduction
Methods
Results and Discussion
References
172
174
177
178
178
178
179
180
181
181
182
183
183
183
190
195
195
201
205
205
207
211
212
212
213
216
217
219
220
221
223
224
227
_~_l_~
_11I__1__
I__LI__YI(PI_^__III ~.---~I~-~ I--~ll-IPI~l~.~-~~~.(1_1
-. -II -^1--11*1~--~-~1~111~. iL-~--XI-W~--~i--~l~ ~--^~-*lpI--~
-~-Plll__~
Appendix B
EVALUATION OF SOLID PHASE EXTRACTION METHODS FOR
SEPARATING DISSOLVED AND COLLOID-ASSOCIATED
CONTAMINANTS IN GROUNDWATER
Abstract
Introduction
Methods
Chemicals
Quantitative Breakthrough of Humic Acid
Reverse Phase Separations
Reverse Phase Separation Systems
Other Colloid Phases
Field Application
Results and Discussion
Evaluation of Humic Acid Passage by Reverse Phase Separation
Systems
Evaluation of Reverse Phase Separation of Colloid-Associated
PAHs
Reverse Phase Separation of Colloid-Associated PAHs in
Groundwater
Conclusions
References
.~-~~--~-~~Y-
228
229
230
232
232
232
233
234
235
235
237
237
239
246
251
252
Appendix C
CALCULATION OF THE EFFECTIVE DIFFUSION COEFFICIENT IN
DOUGLAS FIR
253
Appendix D
ADDITIONAL TIME COURSE PLOTS OF MONOAROMATIC
COMPOUND UPTAKE BY WOOD
258
Appendix E
REPRESENTATIVE GAS CHROMATOGRAMS
264
List of Tables
Chapter 2
Table 2.1. Physical and chemical groundwater parameters from September, 1996.
Table 2.2. Aqueous and tar concentrations of mono- and polycyclic aromatic
hydrocarbons
Table 2.3. Calculated enhancements in polycyclic aromatic hydrocarbon
concentrations at wells W20S and W40M in Sept., 1996.
Table 2.4. Enhancements in polycyclic aromatic hydrocarbon concentrations
in Sept., 1996.
Chapter 3
Table 3.1. Effect of separation methods on the removal of organic colloids from
solution.
Table 3.2. Distribution of pyrene in fractionated W40M groundwater.
Table 3.3. Pyrene fluorescence in W40M groundwater after various treatments
to remove organic colloids.
Table 3.4. Pyrene solubility enhancements by groundwater colloids.
Table 3.5. Pyrene fluorescence quenching by W100S groundwater.
Table 3.6. Organic carbon content of groundwater samples from June 1997.
Chapter 5
Table 5.1. Experimental conditions for sorption isotherms.
Table 5.2. Summary of observed and estimated partition coefficients for
anthropogenic fill materials.
Table 5.3. Elemental composition of organic carbon-containing anthropogenic
fill solids.
Table 5.4. Evaluation of elemental mass balance method for determing
the fractional composition of sorbent mixtures.
47
52
67
72
90
96
100
102
103
105
136
142
155
158
Chapter 6
Table 6.1. Experimental conditions and partition coefficients for equilibrium
Table
Table
Table
Table
isotherms.
6.2. Kinetic uptake of wood particles of various shapes.
6.3. Kinetic uptake by Ponderosa pine chips.
6.4. Kinetic uptake by Douglas fir sticks.
6.5. Estimated characteristic mass transfer times for hindered and
retarded diffusion.
Chapter 7
Table 7.1. Naphthalene retardation factors as a function of depth at Site YYZ.
186
199
200
200
204
215
~--~I~-PL~
1.1~__~_~ .II~I.-~1~-.--.
C1-~LII~---P~~------'~~"~"~C~YIII"^~~''---^----~L--C
-
.IIIYYI-~--
-II~
--
Appendix B
Table B.1. Humic acid passage through solid phase extraction systems.
Table B.2. Polycyclic aromatic hydrocarbon concentrations in tar-equilibrated
water and tar-equilibrated 7 mgc/L Aldrich humic acid solution.
Table B.3. Sep Pak separation of dissolved and colloid-associated polycyclic
aromatic hydrocarbons in tar-equilibrated 18 MO water.
Table B.4. Sep Pak separation of dissolved and colloid-associated polycyclic
aromatic hydrocarbons in tar-equilibrated humic acid solution.
Table B.5. Empore disk separation of dissolved and colloid-associated
polycyclic aromatic hydrocarbons in tar-equilibrated humic acid solution.
Table B.6a. Sep Pak separation of dissolved and colloid-associated polycyclic
aromatic hydrocarbons at monitoring well W20S.
Table B.6b. Sep Pak separation of dissolved and colloid-associated polycyclic
aromatic hydrocarbons at monitoring well W100S.
Table B.6c. Sep Pak separation of dissolved and colloid-associated polycyclic
aromatic hydrocarbons at monitoring well W100M.
238
240
242
243
245
249
249
250
List of Figures
Chapter 1
Figure 1.1. Map of Site YYZ denoting the study region.
Figure 1.2. Material flow diagram for water gas production.
Chapter 2
Sidebar 2.1. Multi-contaminant fingerprint analysis.
Figure 2.1. Map of Site YYZ detailing the field study area.
Figure 2.2. Groundwater sampling apparatus.
Figure 2.3. Groundwater turbidity during continuous slow pumping from
Apr. 9 to 18, 1996.
Figure 2.4. Particle size distributions of groundwater particles collected on
Nuclepore filters.
Figure 2.5. Scanning electron micrograph of filtered groundwater particles
from W40M, Dec., 1995.
Figure 2.6. Representative energy dispersive X-ray spectrum of groundwater
particles.
Figure 2.7. Replicate observations of groundwater mono- and polycyclic
aromatic hydrocarbon concentrations at W40S.
Figure 2.8. Aromatic hydrocarbon concentrations, Dec. 1995.
Figure 2.9. Aromatic hydrocarbon concentrations, Apr. 10, 1996.
Figure 2.10. Aromatic hydrocarbon concentrations, Apr. 18, 1996.
Figure 2.11. Aromatic hydrocarbon concentrations, May, 1996.
Figure 2.12. Aromatic hydrocarbon concentrations, Sept., 1996.
Figure 2.13. Aromatic hydrocarbon concentrations at well W20M as a
function of sample date.
Figure 2.14. Aromatic hydrocarbon concentrations at well W40S as a
function of sample date.
Figure 2.15. Benzo(a)pyrene enhancement factors as a function of turbidity.
Figure 2.16. Stoichiometric electron acceptor requirements for the complete
mineralization of naphthalene.
Chapter 3
Figure 3.1. Stern-Volmer plot of quenched pyrene fluorescence in W40M
groundwater.
20
22
29
32
35
42
44
45
50
54
55
56
57
58
59
62
63
73
78
98
~,
Chapter 4
Figure
Figure
Figure
Figure
Figure
Figure
4.1. Map of Site YYZ detailing the field study area.
4.2. Cross section of the fill material in the field study area at Site YYZ.
4.3. Particle size analysis of anthropogenic fill materials from boring B4.
4.4. Tidal fluctuations on the river and well points.
4.5. Analysis of tidal fluctuations in the anthropogenic fill at Site YYZ.
4.6. Ambient and induced groundwater velocities at the MIT
monitoring well clusters.
Chapter 5
Figure 5.1. Depth profiles of organic carbon, nonaqueous phase liquids and
aromatic hydrocarbons in the B4 boring, 1993.
Figure 5.2. Naphthalene sorption to natural solids.
Figure 5.3. Naphthalene sorption to extracted box waste.
Figure 5.4. Pyrene sorption to coke waste.
Figure 5.5. Sample elemental mass balance calculation to determine the
fractional composition of a sorbent mixture.
Chapter 6
Figure 6.1. Molecular structure of cellulose and lignin polymers.
Figure 6.2. The macrostructure and microstructure of the wood cell wall.
Figure 6.3. Schematic representation of softwood physical structure with
enlarged detail of the interconnecting pit structure.
Figure 6.4. Pictorial representation of Stamm's resistance model for diffusion
through softwoods.
Figure 6.5. Change in aqueous toluene peak area as a function of time for
duplicate flasks containing Ponderosa pine chips.
Figure 6.6. Ponderosa pine sorption isotherms.
Figure 6.7. Douglas fir sorption isotherms.
Figure 6.8. Lignin-octanol linear free energy relationship.
Figure 6.9. Decrease in aqueous peak areas of toluene as a function of time
for varied sizes of Ponderosa pine wood particles.
Figure 6.10. Decrease in aqueous peak areas of toluene as a function of time
for varied sizes of Douglas fir wood particles.
Figure 6.11. Calculation of tangential Douglas fir conductance with Stamm's
resistance model for wood.
~
115
116
118
121
122
125
139
144
148
152
156
168
170
175
176
184
187
188
192
196
197
202
Appendix A
Figure A.1. Comparison of calculated and measured aqueous mono- and
polycyclic aromatic hydrocarbon concentrations in equilibrium with
W40M coal tar.
225
Appendix B
Figure B.1. Polycyclic aromatic hydrocarbon concentrations in groundwater
in June 1997.
247
Appendix D
Figure
Figure
Figure
Figure
Figure
Benzene uptake by Ponderosa pine chips.
O-xylene uptake by Ponderosa pine chips.
Benzene uptake by Douglas fir sticks.
Toluene uptake by Douglas fir sticks.
O-xylene uptake by Douglas fir sticks.
259
260
261
262
263
Appendix E
Figure E.1. W40M coal tar.
Figure E.2. Oil isolated from the B4 boring.
Figure E.3. Pentane/acetone extract from the B4 core.
266
268
270
D.1.
D.2.
D.3.
D.4.
D.5.
------- --- ----------------- --- ----------- --------------
c
~"~"^~-~1~~"1
^------
-- -------- --- ~~~^~"
~~~~'--^~'~-~
-~---------- -r'
Chapter 1.
INTRODUCTION
~-~---
7--
1
I
pi
Introduction
An important class of sites with contaminated subsurface solids and groundwater are
industrial sites with a history of manufacturing operations. These are sites which have been
industrialized for many years, especially before environmental regulations mandated secure
landfill disposal of hazardous wastes, or before it was economical to recycle byproduct
wastes. During this time many waste products may have been buried at these industrial sites,
not only contaminating the subsurface, but also forming the hydrologic units through which
infiltrating precipitation or groundwater moves. Because our understanding of groundwater
transport processes has been developed through study of natural aquifers systems, we pose the
question, "To what extent is our understanding of groundwater transport and reaction
processes applicable to predict the transport of contaminants through these anthropogenic fill
solids at industrial sites?"
This thesis addresses this question in a study to characterize the fate processes acting
upon groundwater contaminants in a water-bearing unit composed of anthropogenic fill at a
former manufactured gas plant site, Site YYZ. (Former manufactured gas plants are also
referred to as "coal tar" sites because of the prevalence of this byproduct in the subsurface
solids.) It is recognized that a unique combination of physical and chemical processes act
upon the groundwater contaminants at this site; however, in a study of 25 former
manufactured gas plant sites, all exhibited similar hydrology and contamination (Luthy et al.,
1994). Over 1000 manufactured gas plants were in operation in the United States prior to
World War II (Environmental Research and Technology Inc and Koppers Company Inc,
1984). Coking plants employed similar production methods, and hence the same
contaminants and waste products may be expected at these sites too. One common
characteristic of manufactured gas plant sites was the use of fill materials to create "made
land" near water bodies (Luthy et al., 1994). There are likely many other industrial sites
which followed this practice. Thus, results from this field study will be applicable at many
sites due to the nature of the contamination or the subsurface solids.
,~II'~'~^~'"--'~~~
"'~xrl~~~.~~1-~~111-- ~--IIIIXI--~._ ---~
1
1
History of Manufactured Gas Production and the Nature of Site Contamination
The history of gas operations at Site YYZ is outlined to provide an understanding of
the reactant, product, and waste materials utilized and produced during gas manufacture.
Reactant and product materials that are used and produced in large quantities during process
operations will tend to compose fill materials and subsurface contamination at industrial sites.
At Site YYZ, the property shoreline was expanded riverward over time by filling in the
wetland with gas manufacture wastes (Figure 1.1). Building rubble from a local fire was also
buried at this site.
Site YYZ is located in the mid-Atlantic region of the United States and operated as a
manufactured gas plant facility from the mid-1800s until 1960. Manufactured gas, also
known as town gas, was initially used for lighting, then later for heating and cooking when
the manufacture process was modified to produce a gas of higher heating value. Natural gas
supplanted manufactured gas as the major gaseous fuel source for heating in the 1950s with
the introduction of interstate gas transmission pipelines. Manufactured gas plants could not
remain competitive with this cheaper alternative gas source and most ceased operation about
this time (Environmental Research and Technology Inc and Koppers Company Inc, 1984). A
liquified natural gas distribution center currently operates at Site YYZ.
Manufactured gas was so named because it was produced from the destructive
distillation of coal. In 1792, William Murdoch distilled coal in an iron retort and produced a
gas for illumination. Within the next thirty years, large scale gas production was refined and
the formation of gas production and distribution companies began throughout Europe and the
Eastern and Midwestern US (American Gas Centenary, 1916).
-9400
a
-9500
I
I
V
I
R2
MIT well cluster
-
*
W20
W40
Recovery well
W100
-9600
-1700
-1600
-1500
-1400
Figure 1.2. Map of Site YYZ detailing the study region. Axes denote distance (ft) from an arbitrary origin.
The approximate locations of the historic shorelines in this landfilled region are noted.
-1300
^-~'C~ -I-~--l-"~L---^l-~ ~^-~--~--~LI--~-~---C-"
-^--~LI"""
---------- ---~---~----------r~-L-_~l.~I-.~-~I---II 11~_
1
Manufactured gas, a mixture of hydrogen, carbon monoxide and low molecular weight
hydrocarbons, was produced by three processes. Coal gas was produced by heating
bituminous coal in a closed vessel (up to 8000 C) until all of the volatile materials were
evolved as gas (Powell, 1945). Water gas, the most prevalent form of manufactured gas, was
produced by alternately reacting coal with air and steam, thus increasing the heat content over
coal gas by the addition of hydrogen. The heating value was further enriched by cracking
petroleum oils in the hot gases (Environmental Research and Technology Inc and Koppers
Company Inc, 1984; Morgan, 1945). Oil gas production was similar to the second step of
water gas production (Environmental Research and Technology Inc and Koppers Company
Inc, 1984). Coal gas was produced at Site YYZ until 1902 when production was switched to
water gas. In 1949 gas production was subsequently converted to oil gas.
The destructive distillation of coal produced many waste products. Water gas
production is summarized in Figure 1.2, explicitly noting the waste streams generated and
their destinations. First, coal was not completely converted to volatile products. In addition
to the gas, light oils, heavy tars, aqueous ammonia solutions (called liquor) and solid char
residues (ash and coke) were also produced during gasification (Environmental Research and
Technology Inc and Koppers Company Inc, 1984; Powell, 1945). The relative abundances of
these gasification products varied somewhat as a function of gasification temperature: gas
production increased from 6 to 18% by weight of coal, and coke production decreased from
81 to 74% with an increase in gasification temperature from 500 to 1100 0 C. Tar, oil and
liquor production remained approximately constant at 5, <1, and 6%, respectively, by weight
of coal gasified (Rhodes, 1945). When total gas production is considered, these amount to
significant quantities of waste from manufacturing gas. In 1939, water gas production in the
US was 4200
x
109 L (Morgan, 1945). Assuming a 12% conversion efficiency of coal to gas
(on a weight basis) and an average gas density of 0.7 g/L (Environmental Research and
Technology Inc and Koppers Company Inc, 1984), 1.3
x
106 metric tons of tar and 21
metric tons of coke were produced as byproducts in this year alone.
x 106
-~I_
Coal
Coke and ash
Oil
Landfilled
Waste heat boiler
.......... Blower
iF
Water
II
Legend
--
Manufactured gas
Tar
I
-1
-
Tar well
- Water/Steam
....... Air or Oil
-
.- Solids
Lagoon disposal
Landfilled
-
i
Sale
I"
Box waste
Ammonia sulphate
~'ur
Sale
Figure 1.2. Material flow diagram for water gas production. Products of manufactured gas
are noted in bold boxes. Adapted from Morgan (1945) and Environmental Research and
Technology Inc. and Koppers Company Inc. (1984).
.--~~ -------------- -c~-- --c-~--^----------~-c----
l--II----- ~-----..~.,~..,
...~~._...
~~xl--------------------- r-"-~-~~~c~ ~~~-~I-~-~~-~--- -r~~ -~~1CI^I--~r
Wastes were also generated from the clean up of the gas stream. The hot gases
exiting the retort contained unwanted impurities of hydrogen sulphide, hydrogen cyanide and
ammonia. Hydrogen sulphide and hydrogen cyanide were removed by passing the gases
through purifier boxes containing a mixture of iron oxides and wood chips. The purifier
boxes were regenerated with air until the build up of ferrocyanide complexes prevented
further use. Ammonia was scrubbed from the gas by passing it through sulphuric acid.
Some of the manufactured gas waste products had commercial value. By the turn of
the century, coal tar had become an important raw material for chemical synthesis (e.g.
naphthalene, tar acids and tar bases) and for the manufacture of creosote and road tars. Spent
oxides or "box waste" could be sold for sulphur recovery (Gollmar, 1945). Ammonia
sulphate from gas scrubbing was feedstock for fertilizer use (Wilson, 1945).
Waste products with no commercial value were land-filled. Until 1900 there was little
market for coal tars, and tars generated up to this point were probably disposed on-site in
sludge pits, tar ponds and disposal wells. Ash wastes had little use and were most likely
landfilled. The use and disposal of light oils and coke are not clear (Environmental Research
and Technology Inc and Koppers Company Inc, 1984).
Groundwater contamination at former manufactured gas plants resulted from the onsite disposal of tars, oxide wastes, ash and coke. Coal tars are viscous dense organic liquids
composed of mono- and polycyclic aromatic hydrocarbons. Many of these compounds are
potential carcinogens and are EPA priority pollutants. Oxide wastes are acidic solid wastes
containing cyanides and heavy metals (Environmental Research and Technology Inc and
Koppers Company Inc, 1984). Ash is primarily composed of aluminum, silicon and calcium
oxides. All of these wastes have been observed to some extent in the subsurface at Site YYZ.
Research Outline
The purpose of this thesis was to understand some of the groundwater fate processes
and characteristics of the fill solids which were unique to this, and other industrial sites, with
water-bearing units composed of anthropogenic fill materials. The fate processes acting upon
the groundwater contaminants at Site YYZ are discussed in Chapter 2. The subsurface source
and fate of dissolved coal tar constituents were assessed by comparing the observed
groundwater concentrations of aromatic hydrocarbons with measured coal tar-equilibrated
aqueous concentrations of compounds with a range of six orders of magnitude in solubility.
Field observations were made under both ambient and induced groundwater gradient
conditions. The physical-chemical characteristics of groundwater colloids which enhanced the
groundwater solubilities of hydrophobic compounds at Site YYZ are described in Chapter 3.
Characteristics of the solids composing the anthropogenic fill at Site YYZ are
summarized in Chapters 4 and 5. Chapter 4 presents some of the issues that may be
important for modelling flow through fill solids. An extensive hydraulic characterization of
these materials was not undertaken as part of this thesis. Chapter 5 presents the physicalchemical properties of the distinct fill materials found at Site YYZ, with respect to their
chemical composition and their capacity to sorb nonpolar organic contaminants.
Wood was recognized as a material that may be buried as fill at many sites with a
history of industrialization. Results from a study of monoaromatic hydrocarbon sorption to
wood are presented in Chapter 6. A free energy relationship to predict wood sorption from
octanol water partition coefficients is also presented.
The experimental results of this thesis are summarized again in Chapter 7 to introduce
a discussion of future research areas which will broaden the understanding of contaminant
transport at industrial sites with hydrologic units composed of anthropogenic fill. Some
general guidelines for remedial investigation of these sites are also suggested.
Two appendices summarize results that were not central to the understanding of
contaminant transport at this industrial site. The applicability of Raoult's Law to predicting
equilibrium dissolution of coal tar is discussed in Appendix A. Appendix B presents a
reverse phase separation method for quantifying in situ colloid-associated contaminants.
--~---^-~-I
-----I-------------~--
-----~ ~
^-~
--"-~-~--'-'----^I-"------I--~---I- ~-----
---- ----'---
-
,---
---- --------
References
Consolidated Gas Electric Light and Power Company of Baltimore (1916). American Gas
Centenary 1816-1916.
Environmental Research and Technology Inc; Koppers Company Inc (1984). Handbook on
Manufactured Gas Plant Sites.
Gollmar, H. A. (1945). "Removal of sulfur compounds from coal gas." In Chemistry of coal
utilization. H. H. Lowry, Ed. New York, John Wiley & Sons, Inc. II: 947-1007.
Luthy, R. G.; Dzombak, D. A.; Peters, C. A.; Roy, S. B.; Ramaswami, A.; Nakles, D. V.;
Nott, B. R. (1994). "Remediating tar-contaminated soils at manufactured gas plant
sites." Environmental Science and Technology 28: 266A-276A.
Morgan, J. J. (1945). "Water gas." In Chemistry of Coal Utilization. H. H. Lowry, Ed. New
York, John Wiley & Sons, Inc. II: 1673-1749.
Powell, A. R. (1945). "Gas from carbonization - Preparation and properties." In Chemistry of
Coal Utilization. H. H. Lowry, Ed. New York, John Wiley & Sons, Inc. II: 921-946.
Rhodes, E. 0. (1945). "The chemical nature of coal tar." In Chemistry of Coal Utilization. H.
H. Lowry, Ed. New York, John Wiley & Sons, Inc. II: 1287-1370.
Wilson, P. J. (1945). "Ammonical liquors." In Chemistry of Coal Utilization. H. H. Lowry,
Ed. New York, John Wiley & Sons, Inc. II: 1371-1481.
Chapter 2.
AROMATIC HYDROCARBONS IN GROUNDWATER AT A COAL TAR SITE
.--------------
~----------- ------- ---- ----"----------------LI~-~ll~c~ ~~IICllllla--
Abstract
The phase transfer and reaction processes acting upon aromatic hydrocarbons in the
groundwater at a coal tar site were hypothesized by comparing the groundwater fingerprint of
individual compound concentrations to measured aqueous concentrations in coal tarequilibrated water. The source of aromatic hydrocarbons in the groundwater was the
dissolution of residual nonaqueous phase liquid coal tar. Dissolution occurred under
equilibrium conditions at this site. Evidence for colloid-enhanced solubilization of polycyclic
aromatic hydrocarbons (PAHs) was found at some monitoring wells. PAH concentrations in
the groundwater were elevated above tar-water equilibrium by sorption to colloid particles and
suspended organic matter. Concentrations of xylenes, naphthalene and methylnaphthalenes
were biologically attenuated at the shallow wells. Sulphate and sulphide were the only redox
couple present in sufficient abundance to account for the loss of aromatic hydrocarbons,
suggesting that sulphate reducers were important degraders at this site. An induced
groundwater gradient had no effect on compound fate processes or groundwater chemistry
compared to ambient gradient conditions.
Introduction
Fingerprinting may be used as a tool for probing groundwater fate processes and the
efficacy of groundwater clean-up approaches (Sidebar 2.1). Site investigations begin with a
hypothesized conceptual model of the governing transport equations for the contaminants
present. This conceptual model, including source and sink terms, may be used to predict the
distribution of contaminants in space and in time. The predicted distribution of compounds is
likely not exact when compared to actual field data; however, if the governing transport
equations are correct, the relative distributions of compounds in the conceptual model will
match the relative distributions in the field data set. When the field data does not match the
conceptualized compound distributions, deviations between the data sets which vary
systematically with compound physical-chemical properties (e.g., octanol-water partition
coefficient, Henry's Law partition coefficient, Hammett constant) may be used to hypothesize
additional fate processes that occur at the field site. Better predictions of remediation
effectiveness or compound transport can be made once these additional processes are included
in the conceptualized site model.
The use of trends in compound physical-chemical properties to investigate groundwater
fate processes is only possible with analyses of individual compounds, and not with bulk
measures of groundwater contamination (e.g., total petroleum hydrocarbons (TPH), volatile
organic compounds). When multi-component plumes exist, one compound may account for
the majority of the mass in a bulk measure. Thus monitoring TPH, for example, only
describes the space or time trends of one, or several, of the most abundant compounds.
Information gained from the relative distributions of less abundant or less soluble plume
constituents would be lost. Monitoring multi-constituent fingerprints may involve more
intensive and expensive data collection and analysis; however, judicious use of these methods
may save on future remediation costs. For example, fingerprinting has been used to identify
natural attenuation of monoaromatic hydrocarbons (Thierrin et al., 1995; Beller et al., 1995)
and halogenated solvents. At these sites implementation of remedial measures may be
unnecessary. The presence of nonaqueous phase liquids (NAPL) was also deduced from the
--
i -- -----IIi---
-------
---- --C ------~---'--- ~'x
"I- ~~~'^~------------
~-1~1~--r----(111 L^II------. ~~-~~-~~--
I
Sidebar 2.1. Multi-component fingerprint analysis.
Fingerprinting is a method by which the relative ratios of organic compound
concentrations in an environmental sample are used to deduce transformation processes
occurring in the environmental system.
Example Application
Assume compounds A, B and C have the same physical-chemical properties (e.g, chemical
formula, octanol-water partition coefficient) and differ only in the degradation rates in
reaction X with C reacting twice as fast as B and four times as fast as A. If the source of
A, B and C had a known composition given by the solid line (below), an environmental
sample with the composition given by the dashed line would suggest that reaction X was
occurring in the environmental system since concentrations of A, B and C are depleted to
increasingly greater extent relative to the known source.
C H RO
ATOGRA PI
C
SEPA RATION
Fingerprinting may be used to deduce groundwater fate processes by gathering data over
time, or space, and comparing relative compound distributions.
Conceptual Model
Since A, B and C have the same physical-chemical properties, the distributions of
compound concentrations relative to the known source concentration should be the same
for each and overlay one another when plotted in time or space.
Actual Distribution
-IrE rTANAP
D
E
Ifthe relative concentration distributions were obtained as shown above, these data would
suggest that reaction X was a groundwater fate process acting on these compounds. Other
data could then be gathered to support this hypothesis (e.g., concurrent appearance of
reaction products).
relative compound distributions in time (MacKay et al., 1996) and space (Jackson and
Mariner, 1995). Site remediation with unamended pump-and-treat could be eliminated from
consideration as a clean-up scheme at these sites due to the long times necessary for
dissolution of the NAPL.
This chapter details the use of fingerprinting to deduce the fate processes affecting
mono- and polycyclic aromatic hydrocarbons (MAHs and PAHs) under ambient and induced
gradient flow at a coal tar site. Residual tar was found in subsurface solids at this site
(Chapter 5). Thus, the source of aromatic hydrocarbons in the groundwater was hypothesized
to be equilibrium dissolution of coal tar. The conceptualized contaminant concentrations were
the aqueous concentrations of aromatic hydrocarbons measured from a batch equilibration of
purified water and site coal tar. The extent to which field measurements of individual
compound concentrations deviated from the conceptual model was quantified by calculating
the ratio of the observed (field) concentration to the measured (batch tar-water equilibration)
concentration. This value was referred to as the enhancement (or depletion) factor when the
observed concentration was greater (or less) than the measured concentration.
Field observations focussed on individual MAH and PAH compounds such that the
range of physical-chemical properties of these contaminants would yield insight to
hypothesize their fates at this site. Comparison of contaminant concentrations and
groundwater quality parameters before and after inducing flow indicated that the gradient had
little effect on aromatic hydrocarbon concentrations. However, several unique physical and
chemical processes were found to act upon these compounds under both ambient and induced
gradient flow. First, coal tar dissolution was hypothesized to occur under equilibrium
conditions at this field site. Secondly, concentrations of hydrophobic polycyclic aromatic
hydrocarbons were present above expected tar-water equilibrium solubilities, suggesting
enhancement by the presence of colloids in the groundwater. Finally, concentrations of
mono- and diaromatic hydrocarbons were biologically attenuated at this site.
-~-'~~'~-"----~
'~~
-'~~--~~--I-------"~--I--^---~'-~~-'
~
I--^-L-'~'X---------
Site Description
The coal tar site (Site YYZ) was located in the mid-Atlantic United States and is a
former manufactured gas plant at which remedial actions are being undertaken to abate
groundwater contamination. Briefly, the shallow unconfined water-bearing unit has been
contaminated by a surface fuel oil spill. Oil percolated through the vadose zone and is now
distributed on the water table. A series of wells have been installed to depress the local water
table and induce flow of the oil to these wells for recovery. Prior to the oil spill, however,
the groundwater and subsurface solids at this site had also been contaminated by wastes from
gas manufacture, primarily coal tar. Therefore, a "pump-and-treat" remediation of the
groundwater and solids at depth within the zone of influence is occurring as a result of the
flow induction. This study focussed upon the changes in chemical and physical processes
acting upon gasification contaminants resulting from this "pump-and-treat", well below the
zone of oil contamination. It was possible to do so because the bulk of the groundwater flow
was at depth in this water-bearing unit (Chapter 4).
The field study at Site YYZ was conducted in a water-bearing unit composed of
anthropogenic fill materials. As operations expanded at the site, the western boundary was
extended by filling in the river as depicted in Figure 2.1. The region of the field study is
located just west of the original predevelopment shoreline in "made land" that was extended
out over the time period to 1890. Ash, slag and other solid wastes (e.g., wood chips) from
gas manufacture were considered suitable fill materials at the time and were used as fill at
Site YYZ. In addition, building debris from a local fire was also used. Evidence of these
materials were noted in drilling logs from the field study region and observed in a boring
obtained during this study (Chapter 5). The silty historic river bottom forms a semi-confining
aquitard at the base of this unconfined water-bearing unit.
The ambient groundwater gradient was estimated from head distributions in fully
penetrating, large diameter (0.3 m) extraction wells (Figure 2.1). The fully screened wells
gave head measurements that were averaged through the depth of the fill material. The wide
well diameters minimized effects of diurnal head variations. A plane was fit through the
reported values and the ambient head gradient was found to be 0.006 in the WSW (254')
direction (SigmaPlot, Jandel Scientific). The ambient pore water velocity was estimated to be
365
2.46'
2.46'
1996
Shoreline
S3.65'
1850'.
Shoreline
1890
Shoreline
-9400 3.48'
2.85'
2.33'
River
3.73'
2.55'
-9500 OB4
R22.50'
•
2.34'
O
-_Ann
I
-1700
MIT well cluster
Recovery well head (ft)
Soil boring
W20
V W40
m
it
VI
2.34'
W100
,
-1600
-1500
-1400
Figure 2.1. Map of Site YYZ detailing the field study area. Axes denote distance (ft) from an arbitrary origin.
The approximate locations of the historic shorelines in this landfilled region are noted. April 10, 1996 ambient
head measurements (ft relative to mean sea level) in the recovery wells are noted at the well location.
-1300
,
--
--------- --- I ----I--------
-- ---
- 11-----
~~~--~1--~1~~1"-----
-I
0.6 to 2 m/d with a porosity of 0.3 and a fill hydraulic conductivity of 15 to 30 m/d
(Chapter 4).
The induced gradient at the MIT monitoring wells was calculated with a conservation
of mass equation. A gradient was induced by pumping extraction well R2 at a rate of
27 L/min from April 12, 1996. Assuming equal radial flow, this volumetric flowrate was
converted to a velocity at a distance, r, from the pumping well by dividing by a cylindrical
area of height saturated thickness, b, and radius r:
v =
27trb0
(1)
where Q (m3/d) was the volumetric flowrate, and b (m) was the saturated zone thickness of
3 m. At the W20, W40, and W100 well clusters, the estimated induced pore water velocities
were 1.2, 0.6 and 0.2 m/d, respectively.
Methods
Well Installation
The effect of an induced gradient on the fate of organic groundwater contaminants was
investigated to the southeast of R2. The monitoring well locations were chosen to minimize
the hydraulic effects of other wells in the recovery system so that R2 induced flow could be
more easily characterized. Additionally, information was known about the fill solids in this
region from the B4 boring. It was thought that there would not be any potentially mobile
coal tar present above residual saturation in the area since tar processing occurred on the
south border of Site YYZ.
A series of multi-level monitoring well clusters were installed at increasing distances
(20, 40, 100 ft) from R2 in Dec. 1994 (Figure 2.1). All wells were 5 cm in diameter and
constructed of stainless steel with 0.6 m screens (0.05 cm slots). Wells were installed using a
hollow stem auger drilled continuously to the well screen depth, using no drilling fluids. The
deepest well was installed first, just above the historic river silt. The medium and shallow
wells were positioned with the screen bottom 0.30 m above the next deepest well. The
shallowest well screen was located at least 1.2 m below the ambient water table and fuel oil
affected capillary zone. Coarse sand was used to backfill the well screens and the fill
material collapsed above the sand. A bentonite seal was installed at the ground surface. No
well development was conducted; however, the first ground water samples were not taken
until Dec. 1995. Subsequent sampling events occurred on April 10 and 18, 1996, and in May
and September, 1996.
Monitoring well nomenclature identifies the distance from the recovery well R2 and
the depth of the well screen. Wells were denoted shallow (S), medium (M) and deep (D).
For example, W40M is the medium depth well located 40 ft from R2. No W20D well exists
as the silty historic river bottom was shallower at this location and only two wells could be
installed with the screen placement criteria.
Groundwater Sampling
Slow pumping methods (Backhus et al., 1993) were utilized to collect groundwater
samples with minimum entrainment of immobile particles. First, wells were "scoped" with
methylene chloride-rinsed aluminum tubing to determine the presence of coal tar in the wells.
Tar depth was quantified according to the height of tar staining on the retrieved tubing after
standing in the well for 10 min. Sample lines were installed with polypropylene and viton
packers (QED, Ann Arbor, MI) to isolate the screened portion of the sampling wells. Sample
line intakes were set at least 0.5 m above the tar when it was present in the bottom of the
well.
Sampling apparatus were allowed to stand overnight to minimize installation
disturbances when pumping was begun. Groundwater samples were collected by peristaltic
pumping at 28 - 34 mL/min. The pump was located downstream of the well and sample
bottle such that groundwater only contacted aluminum tubing or glass (Figure 2.2). Turbidity
was periodically monitored in the field by removing a flow-through cell off line for
measurements with a calibrated turbidimeter (DRT-15CE, HF Scientific, Inc). Conductivity
(HI8333, Hanna Instruments), pH (Orion), and redox potential (platinum electrodes, Orion)
were also monitored in the field. Dissolved oxygen and sulphide were measured by
colorimetric assay (Chemettes, Chemetrics, Calverton, VA). When turbidity levels became
constant, samples for volatile aromatic compounds were collected in 40 mL amber VOA vials
and PAH samples were collected in 2 L amber bottles. PAH samples were spiked at the site
with an internal recovery standard of deuterated phenanthrene in methanol. Methylene
---I
II
I
-~I
-~---~----
--- .1 -------1I~YII ~ 111
1 ^11111_ II____
..C-^l I- C-llllll
-llll~.... _L~-I~I~-C -I-~-- ~_~_l___l~-~r--~~I _~_
Vacuum Gauge
Isolation
Valves
Aluminum
Sampling
Une
to waste
Silicon-stoppered
Flow Through
Monitoring Cell
Peristaltic pump with
variable speed controller
or
Flow Through
Sample Bottle
PVC/Viton
Inflatable
Packer
Stainless Steel
Well Casing
with 0.6m screen
Figure 2.2. Groundwater sampling apparatus.
I~
chloride (100 mL) was then added to the bottles to poison the water and to begin the
extraction process. Groundwater for inorganic analyses was also collected and stored in
60 mL BOD bottles. Filter samples were obtained for microscopic observation of suspended
particles. Small volumes of groundwater (0.5 - 2 mL) were filtered through 25 mm diameter
filters (30 nm pore size, Nuclepore, Pleasanton, CA) in acid-washed Swinnex filter holders
(Millipore, Bedford, MA). RO water (10 mL) was rinsed through the filters, and filters were
stored in a desiccator. BOD bottles and VOA vials were kept refrigerated or in a chilled
cooler until return to the lab.
Chemicals and Glassware
Solvents used for extraction of groundwater and dissolution of compounds were
methylene chloride and methanol (Omnisolve, EM Science, Gibbstown, NJ) and hexane
(Ultra-Resi, J.T. Baker, Phillipsburg, NJ). Recovery standards, internal standards and
quantification standards of deuterated phenanthrene and p-terphenyl (Ultra Scientific, North
Kingstown, RI), 1-bromo-4-fluorobenzene and 1,4-difluorobenzene (Aldrich, Milwaukee, WI)
and m-terphenyl (Ultra Scientific) were used as received. External standards for PAHs were
obtained as EPA 525 PAH Mix A (Supelco, Bellefonte, PA). Standards for quantification of
volatile compounds were made up from neat compounds: benzene, toluene, ethylbenzene,
p-xylene, o-xylene (all ChemService, West Chester, PA), naphthalene (J.T. Baker), 1- and 2methylnaphthalene (Aldrich).
Inorganic compounds included sodium sulphate (Mallinkrodt,
Paris, KY); silica gel (100-200 mesh, EM Science); mercuric chloride (Fluka, Switzerland);
sodium chloride, potassium phosphate monobasic, and sodium fluoride (all Mallinkrodt, Paris,
KY); sodium nitrate and sodium nitrite (Sigma, St. Louis, MO); and 1000 mg/L stock
solutions of iron, aluminum, calcium and silicon (Fisher Scientific, Fairlawn, NJ).
Hydrochloric, phosphoric (Mallinkrodt) and nitric (Ultrex II, J.T. Baker) acids were used for
acidification.
All glassware was soap and water washed, rinsed with reverse osmosis (RO) water and
soaked in chromic/sulphuric acid cleaning solution (Fisher Scientific) for a minimum of 2
hours. Acid-soaked glassware was RO water, methanol, and methylene chloride rinsed.
I---~-C -------~-~~~-I--~-~ '~'~~I
~"~-~c--~~~~- ~""- --- ----~~'I----1
~--?r----------
_
Volatile Compound Analysis
Volatile organic compound samples were analyzed within 1 day upon return to the
laboratory by direct aqueous injection gas chromatography (GC) or by purge and trap/GC.
Cold on-column injections of aqueous samples were made to a Carlo Erba HRGC with a
flame ionization detector (FID) held at 3000 C. A 19 m, 5 jpm film thickness column (0.32
ID, RTX-5, Restek, Bellefonte, PA) was used for chromatographic separation. The
temperature program started at 1030 C and ramped at 80 C/min to 200 0 C. Compounds were
quantified with external standards.
September 1996 samples were analyzed by purge and trap/GC. A 300 PL sample was
injected into 5 mL of RO water and purged for 7 min with helium at 10 mL/min. Purge
gases were concentrated on a Tenax®/silica gel/charcoal trap. The trap was desorbed at
175 0 C for 4 min at a flowrate of 20 mL/min. The desorbed sample was transferred directly to
the head of the GC capillary column through a 0.32 mm internal diameter deactivated fused
silica line held at 150'C. The trap was reconditioned by baking at 2250 C for 5 min.
Compounds were separated with a 60 m, 1 gm film thickness DB5 capillary column (J&W
Scientific, Folsom, CA) and detected by FID with a base temperature of 250 0 C. The
temperature program began at 350 C with a ramp of 100 C/min to 200 0 C and the temperature
was held at 200 0 C for 10 min. Blanks were run between each sample or standard injection.
Internal purge standards of 1-bromo-4-fluorobenzene and 1,4-difluorobenzene were used to
monitor compound recoveries. Purge standards did not vary for blanks, standards or samples,
so no corrections were made to sample concentrations. The coefficient of variation between
all purge standard injections was 10% (n = 13).
PAH Analysis
At the lab, groundwater samples were spiked with an internal recovery standard of
p-terphenyl in methanol without disturbing the methylene chloride layer in the bottle.
Groundwater was first extracted in the sample bottle and poured into a 2 L separatory funnel.
The methylene chloride layer was drained off, and the water was extracted twice more with
80 mL volumes of methylene chloride. One aliquot of methylene chloride was used to rinse
the walls of the empty sample bottle before being added to the separatory funnel. All extracts
for each sample were combined and dried with sodium sulphate which had been baked at
450 0 C for 8 h. The dried extracts were concentrated to about 5 mL with a Kuderna-Danish
concentrator. Extracts (or a subfraction for wells with high compound abundances) were then
transferred into hexane by concentrating under a stream of nitrogen to a final volume of
1 mL.
PAH compounds were separated by silica gel column chromatography. Silica gel was
baked at 450 0 C for 8 h. Fully activated silica gel (2 g) was wet packed and rinsed under
pressure with hexane in glass columns. Groundwater extracts in hexane were applied to the
column and compounds eluted under pressure with the following series: fraction 1: 15 mL
hexane; fraction 2: 9 mL hexane + 5 mL hexane:methylene chloride (8:1); fraction 3: 13 mL
hexane:methylene chloride (8:1) + 2 mL methylene chloride; fraction 4: 7 mL methylene
chloride; fraction 5: 10 mL methylene chloride: methanol (9:1). The PAHs were contained in
fraction 3 which was subsequently concentrated under a gentle stream of nitrogen.
Fraction 3 extracts were analyzed by capillary gas chromatography (Carlo Erba
HRGC). An injection standard of m-terphenyl was added just prior to analysis to quantify the
volume of the extract. A 30 m DB5-MS column (0.32 mm ID, 0.25 jpm film thickness, J&W
Scientific, Folsom, CA) was used for compound separation after cold on-column injection.
The temperature program began at 70 0 C with a ramp of 120 C/min to 120 0 C, followed by a
ramp of 30 C/min to 175'C, and a ramp of 80 C/min to 3000 C and a final hold time of 5 min
at 300 0 C. Compounds were detected by a flame ionization detector and quantified by
measuring peak heights or integrating peak areas and comparing to known external standards.
Phenanthrene and anthracene concentrations were corrected with deuterated phenanthrene
recoveries and high molecular weight PAHs were corrected for recovery with p-terphenyl.
The compound detection limit with this analytical method was a groundwater
concentration of 2 x 106 mg/L for a 2 L groundwater sample, assuming 100% recovery.
P-terphenyl internal standard recoveries averaged 71 ± 20%. This 28% variability in internal
standard recoveries was taken to be an estimate of the analytical variability in compound
concentrations measured in groundwater samples.
-
---------
--- --- ---------------~------ -~"----~-- -"-------
---',---------~I
C-~-~~~~~~---
Verification of Compound Identities
Compound identities were also verified by gas chromatography-mass spectrometry
detection (HP 5995) by matching the temperature program as closely as possible to the above
programs. The mass spectrometer was run with an electron ionization voltage of 70 eV.
Cold on-column injections were made to a DB5-MS column (above). For PAH identification,
a temperature program beginning at 700C, with a ramp of 60 C/min to 300'C and a 15 minute
hold time at 300 0 C was used. Mass-to-charge ratios were collected between 200-350 amu to
increase detection limits of high molecular weight PAHs, integrating for 3 ms and averaging
once. For volatile compounds, 3 mL of groundwater were transferred to a 15 mL vial. The
vials were shaken for 5 min and 1 mL of headspace was withdrawn and injected on the
GC/MS with a temperature program of 350 C for 5 min followed by a ramp at 80 C/min to
200 0 C with a hold time of 15 min. Mass-to-charge ratios of 50-170 amu were monitored.
Tar Analysis
Tar was pumped from W40M during the April 1996 sampling trip after the extraction
well was turned on. A subsample of this tar was removed from the bottle, and a water phase
allowed to separate. The glass tip of a 50 jtL micropipettor (VWR Scientific) was immersed
below the tar-water interface to ensure that only tar was sampled. The outer surface of the
glass tip was wiped free of tar and the tar expelled below the surface of 50 mL of methylene
chloride. This dilution was analyzed for PAHs and volatile aromatic compounds on the Carlo
Erba HRGC with a DB5-MS column. The temperature program began at 350 C with a ramp
of 80 C/min to 3000 C.
Tar-Water Equilibration
Aqueous concentrations of tar components in equilibrium with coal tar were
determined experimentally by mixing 3 mL of tar with 2 L of RO water. Sodium chloride
(1 g/L) was added to match the groundwater conductivity, and mercuric chloride (10 mg/L)
was added to inhibit biodegradation of components. The two phases were mixed with a stir
bar for 2 days and the dispersed tar droplets were allowed to settle for 2 months before
sampling. The aqueous phase was carefully siphoned into a separatory funnel using
aluminum tubing primed with RO water. The aqueous phase was spiked with deuterated
phenanthrene and p-terphenyl. The equilibrated water was extracted three times with
methylene chloride. The solvent extracts were combined and analyzed as described for PAHs.
A small aliquot of tar-equilibrated water was also removed for purge-and-trap analysis of
volatile aromatic compounds, as described.
Inorganic Compounds
Levels of inorganic anions in the groundwater were determined by ion chromatography
(Dionex Ion Chromatograph 16). A 1 mL sample was delivered to the AS4A-SC column
(Dionex, ) and eluted with 0.003 M sodium bicarbonate/0.0024 M sodium carbonate buffer at
a rate of 2 mL/min. Ions were quantified with external standards.
Groundwater cation concentrations were determined by graphite furnace atomic
absorption spectrophotometry (Perkin Elmer 4100ZL). Standards were made up in Q-water
(Millipore, Bedford, MA).
Carbon Analysis
Alkalinity titrations were performed by Gran titration with 0.02 N hydrochloric acid
and an Orion pH meter. Total inorganic carbon was calculated assuming all the alkalinity to
be bicarbonate ions and applying equilibrium dissociation constants (Morel and Hering, 1993).
Total non-purgeable organic carbon was determined by high temperature oxidation
(Shimadzu, 8 gL syringe). Samples were acidified with phosphoric acid and purged with
argon for 10 to 15 min prior to analysis. Purgeable organic carbon was determined by
integrating the purge-and-trap chromatogram, using a benzene response factor (ng/area) for all
peaks present except the naphthalenes for which the naphthalene response factor was used,
due to the inefficient naphthalene stripping.
Surface Tension
Groundwater surface tension measurements were made by the falling drop method
(Harkins and Brown, 1919). Measurements of RO water surface tension were made to verify
this method. A value of 72.2 ± 0.1 dyne/cm (n = 5) was calculated for RO water and
compares with the reported value of 72 dyne/cm at 25 0 C (CRC Handbook of Chemistry and
Physics, 1989).
--
---^-l-~L -~-1II1
---
' -r~-_Il ._.--r.-----l--(------ ---II~----~rX~C-.
.-~---~--~--Ir~--~*ll-~---I~
I~-
Electron Microscopy
Groundwater filters were observed by scanning electron microscopy (Cambridge
Instruments). Particle composition was determined by energy dispersive X-ray analysis (Link
Analytical) of carbon-coated filters. Particle size distributions were calculated from
measurements of particle diameters of particles at random locations on the filter. Due to the
time consuming nature of these manual measurements, only 30 particles were measured on
each filter.
Results
Method Evaluation/Sample Quality
Groundwater sample quality was first evaluated to determine that artifacts from the
groundwater sampling procedure were minimized. No conclusions can be drawn about the
groundwater fate of aromatic hydrocarbons if representative samples were not obtained. It is
generally thought that groundwater from the aquifer unit is being sampled after turbidity
measurements have reached asymptotic levels (Backhus et al., 1993). Groundwater turbidities
were monitored at wells W20S, W20M, W40S and W40M through 5 to 9 days of continuous
slow pumping over April 9-18, 1996 (Figure 2.3) to obtain data about turbidity variations at
these wells.
Asymptotic turbidity levels at all of the wells sampled in April 1996 appeared to be
less than 2 NTU. Turbidities fell within this range for these wells at all other sampling times,
except at W40M in Dec. 1995. At that time it was thought that the sampling lines would
freeze if pumping was continued overnight to allow turbidity stabilization. The last turbidity
measurement before groundwater samples were obtained from W40M that day was 28 NTU.
The long term asymptotic turbidity of groundwater at the W100 cluster is not known, but at
each sampling time sufficient groundwater was pumped to reach asymptotic turbidities over a
12 to 24 h timescale. If the data collected for the W20 and W40 well clusters was
representative of the W100 cluster, the turbidity levels approached on the timescale of a day
were likely the same as those that would have been approached after longer term (i.e.,
weekly) pumping.
~
Elapsed Pumping Time (h)
0
50
100
150
40
0
50
100
W20S
30
R2 on: 1.5 WV
1WV = 1230 mL
20
packer failure
8
6
H
0
50
100
150 200 250 300
0
50
100
50
100
150
200
H0
0
50
100
150
200
0
50
100
150
150
200
200
250
Well Volumes
Figure 2.3. Groundwater turbidity during continuous slow pumping from Apr. 9 to 18, 1996.
Turbidities are plotted as functions of both well volumes turned over (lower x-axis) and
elapsed pumping time (upper x-axis). The initiation of induced gradient flow is noted on
each figure. Instrument readings of turbidities below 10 NTU generally varied by ± 0.2 NTU.
---r-- - -~--r.~.-..l.-.
~1-~----^1_
--1--
-- 111--1-1^~11
~lg__
.-1~-11~---.~-~1_~
_I--~_^l_
L
ll_ __l_ ~I~ _.--_.-..--_-
The size distribution of particles in a groundwater sample also provides an indication
of sampling artifacts. Slow pumping was employed to minimize shearing of the aquifer
formation. High flowrates (i.e., shear rates) cause the release of particles from the aquifer
solids that would be immobile under the ambient groundwater gradient. If the fate of particle
reactive or hydrophobic compounds, such as polycyclic aromatic hydrocarbons, are of interest,
a groundwater sample obtained under high shear conditions may contain compound species
which were not mobile under the ambient groundwater gradient.
Groundwater particles obtained on filters from Site YYZ were generally less than
1 pjm in diameter (Figure 2.4). Thus most of the particles in the groundwater samples were
likely mobile. Filter samples obtained after several hours of pumping wells in Dec. 1995 did
contain larger particles, up to 40 jim in diameter (Figure 2.5). Energy dispersive X-ray
analysis of these particles showed only a silicon peak. Dec. 1995 was the first time that the
monitoring wells were sampled so these particles were likely dislodged from the surrounding
sand pack during sample line installation. There was also concern at this date that sample
lines might freeze so the packers were not allowed to stand overnight before sampling. These
large particles were not observed at the other sample dates and it may be that well disturbance
artifacts were allowed settle by setting packers the night prior to sampling.
The sampling distribution of the mean particle sizes were calculated from the
frequency distributions for a population with an unknown variance. The mean diameter in
Dec. 1995 was between 0.5 and 1.1 pm with a probability of 90%. With the same
probability, the mean particle diameter was between 0.5 and 0.8 jpm at W40S in Dec. 1995
and between 0.8 and 1.4 gm at W20M in Sept. 1996. As the cumulative distribution suggests
(Figure 5.4), there was no significant difference between groundwater particles at W20M and
W40S in Dec. 1995. After pumping was initiated at R2, the cumulative particle size
distribution at W20M shifted to larger diameters; however, the range of mean particle
diameter in Sept. 1996 overlapped the range of mean particle diameter computed for the prepumping Dec. 1995 sample date.
I
I
W40S Dec. '95
W20S Dec. '95
W20S Sept. '96
oo--
-ro
coo
I
-
066
0
00
60
0
00
-Nm
tv
Particle Diameter (ptm)
100
o
-
c'
r0'-
\O
r'
0
C
0
0
0
0
Particle Diameter (pm)
Figure 2.4. Particle size distributions of groundwater particles collected on Nuclepore
filters. Particles were counted by scanning electron microscopy.
*
,:
I£ a.
i
ii'
r
lOOs
10 siiPrJ
I
r
-
L~i3
tI
F~:1
Sfii
rirI
L,~i~-Ths RdainlL'
tI
i.
1
f
S
!
Reraining:
W
et:°"O qos
a
iC
'.6
.:. 1,
F'S=5!1 1
880
O2.
OS=
, -
e2
ch
1i5=
55
Figure 2.5. Scanning electron micrographs of groundwater particles.
tsi
Groundwater Quality Parameters
Water quality parameters (pH, conductivity, redox potential) showed little change over
the entire sampling period from December 1995 until September 1996. Representative values
are reported from Sept. 1996 (Table 2.1). The groundwater was mildly acidic and was
reducing. The shallow wells consistently had the lowest redox potentials at each cluster and
the presence of hydrogen sulphide was noted in winter, spring and fall.
Groundwater turbidity levels at Site YYZ were generally low, but varied over time and
location. As noted, the asymptotic turbidity levels at all of these wells appear to be less than
2 NTU; however, turbidities did show considerable variability over time (e.g., W20M, W40M,
Figure 2.3). The timescales for these transient bursts to decline back to asymptotic levels
may be shorter than suggested by Figure 2.3 because turbidity measurements were made only
periodically. The increased turbidities at W20S (20 h) and W40S (170 h) likely resulted from
packer failures. The integrity of the W20S packer was checked immediately after the
12 NTU value was reported. At this time the packer was loose in the well, suggesting that
the higher turbidity may have been generated in the well with the expulsion of pressurized air
upon packer failure. The packer in W40S was also loose at the end of sampling, although it
is not known when over the last 100 h that it failed. (Failed packers were left in place
because their large diameter still limited exchange of fluids between the screened interval and
the standing water above in the well casing.) All other turbidity spikes were not correlated
with known turbidity release events, and thus represent the natural variability of turbidity
levels at this site.
High values of inorganic and organic carbon were found at all monitoring wells. The
high alkalinity values and acidic pHs indicated very elevated dissolved carbon dioxide in the
groundwater. Up to one third of the total organic carbon in the groundwater was contributed
by volatile coal tar constituents. Of the remaining non-purgeable organic carbon, only about
2 mg/L was chromatographable with the temperature program used for PAH analysis. The
residue remaining after evaporating an aliquot of the methylene chloride groundwater extract
was less than 2 mg/L. Thus, the bulk of the nonvolatile organic carbon was not organic
solvent extractable and is presumably composed of fulvic and humic acids.
-------- P------------------
.r-------- I--I1----------
- '"II~~' --
--g----- 1
-------
x~---~I I'~~I-~--I------- ---~~-p -------------
~~---- ~~--
----
Table 2.1. Physical and chemical groundwater parameters from September, 1996.
W20M
W20S
W40M
W40S
W100D
W100M
W100S
5.4
5.5
5.6
5.4
5.3
5.5
5.6
1.4
(78*)
2.3
0.5
1.6
7.2
7.8
4.3
Conductivity
(field)
(mS)
1.95
1.37
1.76
1.32
1.50
2.23
2.44
Redox Potential
(field)
(mV, H' scale)
-51
-82
-31
-110
-16
+29
-76
Dissolved
Oxygen (IM)
<1
<0.3
<0.3
6
<6
5
<0.3
<80
1600
<80
1600
<80
<80
800
2000
300
<1
1400
NA
20
1400
4000
<1000
<300
<2100
NA
100
<2700
19
18
100
<3
NA
<3
<5
Alkalinity
(meq/L)
27.2
15.7
19.2
14.6
15.4
18.6
15 1
(calc'd)
(atm)
10+084
10+051
10 + 5
10+057
10+07
10+058
10+039
Nonpurgeable
organic carbon
(mg/L)
45
34
34
28
33
33
40
Purgeable
organic carbon
(mg/L)
13
3
16
4
16
14
17
pH
Turbidity (field)
(NTU)
(*prior to 2L
sample)
S2- ol
(M)
SO042
(jM)
NO
3-
(PM)
PO 4 3
(M)
CO
2
NA - not analyzed
1
31
Table 2.1 (cont.). Physical and chemical groundwater parameters from September, 1996.
AlToW
W20M
W20S
W40M
W40S
W100D
W100M
W100S
30
20
4
7
17
31
80
1
1.4
2
1
9
2
10
2400
7200
5000
5500
4200
4700
5500
13
37
40
5
100
370
83
3
70
3
3
12
4
7
NA
NA
1500
NA
NA
NA
1600
NA
NA
1000
NA
NA
NA
1300
(PM)
ADIssolved
(PM)
CaToW
(PM)
FeTo
(PM)
FeDssolved
(IM)
SiTota I
(M)
SiDssolved
(PM)
NA - not analyzed
-----
--
-------
---- --~-'^-~
L---------~s~-------- ----1--"~^"-s"ll--- rc
Il.-.~;.~-- - p---------- -------------
Mineral Phases
The composition of filterable material at Site YYZ could not be determined by energy
dispersive X-ray (EDX) analysis of these filters. A representative EDX spectrum is shown
and did not show any elemental peaks beyond the background from the bare membrane filter
(Figure 2.6) The sparse abundance of material collected on groundwater filters may have
been insufficient to determine submicron particle composition by EDX analysis. It is also
possible that filterable material was organic in composition and thus gave no signal with a
beryllium EDX window. High carbon content (50-90% by weight) coke was observed in the
soil boring (Chapter 5). Perhaps coke fines are mobile in the groundwater but give no
detectable EDX signal due to their virtually pure carbon content.
MINEQL (Schecher and McAvoy, 1994) calculations were made to determine the
possible solids present in the groundwater at each well with groundwater compositions given
by the total species measured and reported in Table 2.1. These calculations indicated that
under equilibrium conditions the only solids present would be gibbsite from the precipitation
of aluminum. Indeed, dissolved aluminum concentrations determined after sample filtration
were less than total aluminum concentrations (Table 2.1).
--------------------p
IX-RRY
Live
100s Preset:
1tOO
Remaining:
Os
eF S
,
L4
5f5f0
iFS=127,
,EMIT
kp
.6
-
8 .ct
ch. 2835.=
....
°"
,
Figure 2.6. Representative energy dispersive X-ray spectrum of groundwater particles.
The spectral locations of the primary calcium, aluminum, silicon and iron peaks are noted.
50
~/CI~^I~ _
r~-iPIllpll.l.~~--l- ~
IIYII-~-~CI
LII^..
-L Ip--~i. __ I.--_I.-XI~PIUP_1~I.--I1~-IXI( --Lli. _ ____~
~I~IY~C-l
_.II_
)-----Ls~s~s~
^III_________L_~R__I__-~l^
Aromatic Hydrocarbons in Groundwater
Concentrations of aromatic hydrocarbons in coal tar and coal tar-equilibrated water are
reported in Table 2.2. Numerical values of aromatic hydrocarbon concentrations at W40M
and W40S are also included in this table. Values for W40S are the average + the standard
deviation of three groundwater samples obtained over a 36 hour period.
Groundwater concentrations of aromatic hydrocarbons were graphically compared to
measured batch equilibrium aqueous concentrations. An enlarged sample plot is shown for
W40S (Figure 2.7). Here data are plotted with bidirectional error bars representative of the
28% variability in internal standard recoveries of p-terphenyl. Data from 3 samples taken
over a 36 h time period prior to extraction well pumping were included to demonstrate
variability between concentrations in replicate samples. In addition to the experimental data,
lines denoting agreement of observed concentrations with tar-water equilibrium (1:1) and
observed concentrations which differ by a factor of 2 (2:1 and 0.5:1) from tar-water
equilibrium are also shown. The range of concentrations of most of the compounds generally
fell within the 2:1 and 0.5:1 lines. (The low volatile organic compound concentrations are
addressed later in the discussion section.) If the variability among individual samples taken
over a short time period at W40S is representative of the heterogeneity of samples obtained at
the other wells, then compound concentrations which differ from measured tar-water
equilibrium by a factor of 2 are likely not significantly different than tar-water equilibrium
levels.
The comparisons of observed groundwater concentrations to measured batch
equilibrium concentrations at all sample dates are shown in Figures 2.8 to 2.12. One-to-one
lines are also plotted for comparison with the equilibrium dissolution case. Points which fall
below the 1:1 line are depleted with respect to equilibrium tar dissolution and points which
fall above are enhanced over solubility in the presence of tar. For clarity, error bars were not
included in these figures.
Table 2.2. Aqueous and tar concentrations of mono- and polycyclic aromatic hydrocarbons.
Subcooled liquid solubilities are from Miller et al., 1985, except for 2-methylnaphthalene, chrysene, and
benzo(ghi)perylene which were calculated from the Yalkowsky equation (Schwarzenbach et al., 1993). W40S groundwater
concentrations are the average and standard deviation of 3 samples obtained over a 36 hour period.
Compound
and Figure
Abbreviation
Tar
Concentration
(mg/L)
Calculated
Equilibrium
Aqueous
Cone. (mg/L)
Measured
Equilibrium
Aqueous
Cone. (mg/L)
W40S
Apr. 10, 1996
Groundwater
Cone. (mg/L)
W40M
Apr.10, 1996
Groundwater
Cone. (mg/L)
benzene
1.1
3 ± 0.02
5.4
toluene
0.02
< 0.1
< 0.05
ethylbenzene
4900
1.36
1.21
1.4 ± 0.1
0.5
m, p-xylene
MX
990
0.32
0.28
< 0.05
0.4
o-xylene
OX
3200
1.0
0.87
0.073 ± 0.006
1.1
naphthalene
NA
55 700
7.8
10.5
0.4 ± 0.1
4.8
2-methylnaphthalene
2MN
29 000
1.0
2.09
<1
3
1-methylnaphthalene
IMN
23 700
0.76
1.46
< 1
4
PH
19 700
0.11
0.073
0.063 - 0.0007
0.12
phenanthrene
2-methylphenanthrene
anthracene
3900
AN
5000
0.027
0.012
0.011 i 0.001
0.02
fluoranthene
FL
6500
8.7 x 10-3
3.3 x 10' 3
3.8 x 10"3 : 5 x 104
0.015
pyrene
PY
9300
6.4 x 10"
1.4 x 10' 3
2.5 x 10-3
5 x 10-4
0.011
Table 2.2 (cont.). Aqueous and tar concentrations of mono- and polycyclic aromatic hydrocarbons.
Compound
and Figure
Abbreviation
Tar
Concentration
(mg/L)
Calculated
Equilibrium
Aqueous
Cone. (mg/L)
Measured
Equilibrium
Aqueous
Cone. (mg/L)
W40S
Apr. 10, 1996
Groundwater
Cone. (mg/L)
W40M
Apr.10, 1996
Groundwater
Conc. (mg/L)
benz(a)anthracene
BA
3900
6.9 x 104
7.7 x 104
3.8 x 104 L 5 x 10"-
7.9 x 10.
chrysene
CH
3600
9.6 x 10'
6.8 x 104
4.2 x 10
3 x 10's
6.3 x 10.
1.8 x 104 k 5 x 10 s
3.4 x 10*
1.4 x 10" 1 1.4 x 10.3
3.4 x 10.
benzo(b&k)fluoranthene
4300
2.4 x 10"
benzo(e)pyrene
3700
1.9
benzo(a)pyrene
indeno(123-cd)pyrene
benzo(ghi)perylene
BaP
3600
IP
1200
BP
1200
1.1
x
104
6 x 10-s
x
104
4-
5.8 x 10.
3.5 x 104
1.7 x 104
6.2 x 10"'
7 x 10-s + 4 x 10"
2 x 10.
1.6 x 104
6
10"
2 x 10.
x
10"s
+
9
3
x
x
104
le+2
le+1
le+0
le-1
le-2
le-3
le-4
le-5 wle-5
le-4
le-3
le-2
le-1
le+0
le+1
le+2
Measured Equilibrium Concentration (mg/L)
Figure 2.7. Replicate observations of groundwater mono- and polycyclic aromatic
hydrocarbon concentrations at W40S. Observed concentrations are plotted against
measured aqueous concentrations in equilibrium with W40M coal tar. Error bars represent
analytical variability and are smaller than the data symbols where not visible. Compound
identities are aligned above or below the respective data points and are abbreviated in
Table 2.2.
I
I
I
I
W20S
W20M
le+1
-.
.
le+1
le+O
le-1
le-1
le-2
le-2
le-3
le-3
le-4
le-4
le-5
le-5
'""
" . "
W40M
le+
le+O
le+O
oC4
le-2
0
le-3
*
B
le-3
le-4
le-5
e
S'
©n
W100D
,
=
-
0
W100M
W 100S
le+1
le+1
le+O
le+0
le-I
le-1
le-2
le-2
le-3
le-3
le-3
le-4
le-4
le-4
le-5
le+O
1
le-1
.
I
.,=
)t 1
+
+
t
le-2
.~
.r
*O
le-5
"a
le-5
C
,I
TT
le-2
le-5
',dI
.
le-1
H
le-4
r
..
W40S
le+1
le-1
"...
+04)4
.
..
*
4)
I
)
" . "
I
I
I
4
4)
4)
+
)
Equilibrium Concentration (mg/L)
Figure 2.8. Aromatic hydrocarbon groundwater concentrations, Dec. 1995. Concentrations
below detection limits are noted with an arrow and a symbol at the detection limit.
4
W20M
le+l
le+O
le-1
le-2
le-3
le-4
le-5
I
I
I
I
I
+
+
W40M
le+1
..... , .
n.
, .
W40S
, . ..
..
..
le+O
le+0
le-1
*H
le-2
o
le+1
le-1
le-2
le-2
BaP
(U
"0
O~~~-,
le-3
le-3
le-4
le-4
le-5
-- J--......
......
le-5
Equilibrium Concentration (mg/L)
Figure 2.9. Aromatic hydrocarbon groundwater concentrations, Apr. 10, 1996.
Concentrations below detection limits are noted with an arrow and a symbol at the
detection limit.
..
------------- L--
I
W20M
1le+l r -,.
le+1
le+O
le+O
le-1
le-1
le-2 r
le-2 r
le-3
le-3
le-4 t
le-4
I' .I...
" . "I. "I. " ... 1
lel+1
o
W40S
-n
,
le.
+
le+O
P
le-1
le-2
le-3
. .....
le+O
le-1
C0
6+le-5
W40M
o
_~~
W20S
B
le-5
I
le-2
J
le-3
le-4
le-4
le-5
le-5
,
t
-
"
......
L
6
6
O
Equilibrium Concentration (mg/L)
Figure 2.10. Aromatic hydrocarbon groundwater concentrations, Apr. 18, 1996.
Concentrations below detection limits are noted with an arrow and a symbol at the
detection limit.
Ui -
+0
W20M
le+1
le+O
le-1
le-2
le-3
le-4
........
le-5
W40M
W40S
le+ l .....
.....
......
,1
...........
le+..
le+O
le+0O
le-1
le-1
PH
le-2
0
le-3
"7
le-4
le-2
/
le-5
aP
.
le-3
le-4
"le-5
....
a . ..
W100M
le+ 1 ....,
le+O
W100S
, . ,. .,. .....
*
le+O
le-1
le-1
le-2 r
"
le-2
le-3
-
le-3
le-4
le-5
. ......
le+ I
,-
le-4
. ." u"I
........
-4..
I
le-5
I
.
I
Equilibrium Concentration (mg/L)
Figure 2.11. Aromatic hydrocarbon groundwater concentrations, May, 1996.
Concentrations below detection limits are noted with an arrow and a symbol at the
detection limit.
+ +
..I
--
--- ------------------------- I'--- --------~--U-----~-- - -----~-~ ~lpll~----'~----r
-.-.- ------''~T~-~-~~-~ ~~~~-1~~1~111111111111111
1
W20S
W20M
. ......
le+l r
le+0
le+0
le-1
le-1
le-2
le-2
le-3
le-3
le-4
le-4
le-5
le-5
*
W40M
1 e-2
le-3
.
W40S
le+1
"
le+O
le+O
le+0O .
le-1
le-1
B
le-2
le-2
e
le-3
le-3
0
0
v
le-4
+
4
le-4
*
le-5
...
t
O
W100M
W100D
W100S
le+1
le+1"
le+O
le+O
le+O
le-1
le-2
le-1
le-1
0
le-2
le-2
le-3
le-4
le-3
le-4
le-4
le-5
.
le-5
.....
le-5
le-3
4)~m4)4
-
-
4 4 4 -
14
-
1-*
I
I
)4
I
I
I
l
'rld
*
I
I
I
Equilibrium Concentration (mg/L)
Figure 2.12. Aromatic hydrocarbon groundwater concentrations, Sept., 1996.
Concentrations below detection limits are noted with an arrow and a symbol at the
detection limit.
0
I
+
Discussion
Changes in Groundwater Chemistry from an Induced Groundwater Gradient
Pumping of the extraction well, R2, increased the groundwater hydraulic gradient at
Site YYZ above the ambient level. Over a 4 hour period after starting R2, the vacuum
required to peristaltic pump groundwater at the W20 and W40 well clusters increased,
indicating a decline in the watertable. By Apr. 18, groundwater levels had decreased about
0.3 m from the pre-pump levels at the W20 cluster and about 0.15 m at the W40 cluster. The
induced gradient was linearly approximated to be 0.025 between the clusters.
Turbidity is likely the first groundwater quality parameter to show a response to the
induced gradient. A turbidity pulse may be released in response to the increased shear from
the higher velocities created by pumping. If the composition of fill solids at Site YYZ is
similar at all of the well clusters, the greatest turbidity release would be expected at W20
since the local groundwater velocity increased by a factor of 3 with the onset of pumping.
Sampling at W20S was initiated coincident with the start-up of R2 and showed turbidity
levels less than 2 NTU (Figure 2.3). There appeared to be a slight increase in turbidity
values at W20M after R2 began pumping (Figure 2.3). Turbidity increased from 0.65 NTU to
0.95 and 1.25 NTU after 5 and 7 hours of pumping, respectively; however, these fluctuations
were less than the levels observed during pre-pumping monitoring and were not maintained
with continued pumping. Similar trends were observed at W40S and W40M in response to
pumping from R2 pumping. (Groundwater sampling was suspended at W40M for 3 days to
remove tar from the bottom of this well when R2 was started.). Thus the induced gradient
did not cause sustained turbidity releases from the fill solids, relative to pre-pumping levels.
Particle size distributions of suspended material in Site YYZ groundwater also did not
appear to change as a result of the induced gradient. Particle sizes were determined by
microscopic observation of groundwater filters. Problems with insufficient sample volume
and incomplete salt removal limited the comparisons that could be made by this method with
filters from the April 1996 sampling event. Filters from Dec. 1995 (pre-pumping) and Sept.
1996 (pumping) sampling events were compared instead. The W20 cluster was expected to
show the greatest change in groundwater velocity as a result of pumping. Little difference
was seen between particle sizes at W20S at these two dates or between W20S and W40S
;_____~~__~ __~~~_ ~~~..___~_I~. 1. i ......~
-------
I-- ~--'~~~-'-~"I'~'-----\---~-~cx--^'--I- ~--~--~'~-~-~'-~' ~----'llssYIIIIIIII~LIII-----------~~
(Figure 2.4). Due to the time consuming nature of the microscopic counting, no replicate
measurements were made to quantify the statistical spread of these distributions. The less
than 2 NTU turbidity measurements made at these wells at sampling times subsequent to R2
pumping are likely stronger evidence that the induced gradient had little effect on the mobile
particle load in Site YYZ groundwater.
There were no differences in groundwater quality parameters between pumping and
pre-pumping sample dates. Changes in groundwater quality parameters may indicate that the
upgradient "origin" of groundwater at a monitoring well had changed as a result of the
induced gradient. This does not appear to be occurring at Site YYZ.
Groundwater concentrations of selected aromatic hydrocarbons were plotted as a
function of time to determine if their levels had changed in response to pumping. Sample
plots are shown in Figures 2.13 and 2.14. R2 pumping was initiated between Apr. 9 and
Apr. 18, but no systematic trends in compound concentrations were observed between prepumping and pumping sample dates. For example, pyrene concentrations declined with time
at W20M and W40S, but naphthalene concentrations increased (W20M) or remained constant
(W40S). Possible phenomena which might result in changes in contaminant concentrations in
response to an induced gradient are rate-limited solid-water mass transfer, or an imbalance
between compound biodegradation and advection rates. No processes such as these appeared
to affect aromatic hydrocarbon concentrations in the groundwater at Site YYZ up to 5 months
after pumping was initiated.
The groundwater fate of aromatic coal tar constituents at Site YYZ was investigated
with no differentiation between fates under ambient and induced gradient conditions since no
apparent changes in groundwater quality parameters or aromatic hydrocarbon concentrations
were observed between pre-pumping and pumping sample dates.
0.10
14
Naphthalene
12
0.09
Phenanthrene
0.08
10
0.07
8
0.06
0.05
0.04
0.03
0.02
Dec 13'95
Apr 9 '96 Apr 18'96 May 29'96 Sep 25'96
Dec 13'95
Apr 9 '96 Apr 18 '96 May 29'96 Sep 25'96
Dec 13'95
Apr 9 '96 Apr 18'96 May 29'96 Sep 25'96
...
..
0.0009
0.0035
0.0008
0.0030
0.0007
0.0025
0.0006
0.0005
0.0020
0.0004
0.0015
0.0003
0.0010
0.0002
0.0001
0.0005
Dec 13'95
Apr 9 '96 Apr 18'96 May 29'96 Sep 25'96
0.00025
0 00050
Benzo(ghi)perylene
0.00045
0.00020
0.00040
..
..
..
...
..
..
..
..
..
..
..
..
..
..
..
.. ..
..
..
..
..
..........
0.00035
0.00015
0.00030
0.00025
0.00010
0.00020
0.00015
0.00005
0.00010
0.00005
0.00000
Dec 13'95
Apr 9'96 Apr 18'96 May 29'96 Sep 25'96
Dec 13'95
Apr 9 '96 Apr 18 '96 May 29'96 Sep 25'96
Figure 2.13. Aromatic hydrocarbon concentrations at well W20M as a function of sample
date. The error bars denote the analytical variability in the individual measurements. For
comparison, the measured tar-water equilibrium concentration is denoted as a solid line in
each plot. The dotted lines denote the analytical variability of the tar-water equilibrium
concentrations.
-- ^lllrll
.L^-----~-^
------II -~LII_
I~-~-.P~--.II~CI
--L--LII~1___11 ~-_I^~-~--I--IC_--
-'"LI----~ --l ----------
0.12
0.52
N
0.50
Phenanthrene
0.11
0.48
0.10
0.09
0.46
0.08 -
0.44 -
0.07
0.42 -
0.06
0.05
SO
0.010
{
0.04
0.40
0.38
I-C~-~~
0.03
1
Dec 13'95 Apr 9 '96 Apr 18 '96 May 29'96 Sep 25'96
Dec 13'95 Apr 9 '96 Apr 18 '96 May 29'96 Sep 25'96
0.0009
a
..
..
....
.. .
...
. ... .. .. .... ..
Pyrene
0.009
() isr
0.0008
. I...
0.008
0.0007
0.007
0.006
0.0006 -
0.005
S.
0.004
0.003
0.002
0.0005
......................
............
. . . .
0.0004
...............
..
..............
0.0003
0.001
Dec 13'95 Apr 9 '96 Apr 18 '96 May 29'96 Sep 25'96
Benzo(a)pyre
0.00035
Dec 13'95 Apr 9 '96 Apr 18'96 May 29'96 Sep 25'96
i
0.00025
, 0.00020
Benzo(ghi)perylene
......................................
0.00030
0.00015
0.00025
j
0.00020
S0.00010
0.00010
I
0.00005 F
Dec 13'95 Apr 9 '96 Apr 18 '96 May 29'96 Sep 25'96
0.00005
0.00010
0.00005
0.00000
Dec 13'95 Apr 9 '96 Apr 18 '96 May 29'96 Sep 25'996
Figure 2.14. Aromatic hydrocarbon concentrations at well W40S as a function of sample
For
date. The error bars denote the analytical variability in the individual measurements.
comparison, the measured tar-water equilibrium concentration is denoted as a solid line in
each plot. The dotted lines denote the analytical variability of the tar-water equilibrium
concentrations.
I
Fate of Coal Tar Constituents at Site YYZ
Equilibrium Coal Tar Dissolution
Aqueous concentrations of mono- and polycyclic aromatic hydrocarbons indicate that
the groundwater at Site YYZ was in equilibrium with coal tar at the outset of this study.
Over a range of 6 orders of magnitude in compound aqueous solubility, aromatic hydrocarbon
concentrations generally fell on the 1:1 line when plotted against the corresponding
laboratory-measured concentrations of tar-equilibrated water (Figures 2.8 to 2.12).
(Deviations from the 1:1 line greater than can be accounted for with analytical error are
discussed later.) It was only by observing such a wide range of compound solubilities that
this equilibrium conclusion could be made. Use of volatile compounds alone may not have
indicated groundwater-tar equilibrium at Site YYZ. Groundwater concentrations of
monoaromatic hydrocarbons and naphthalenes at some of the wells (W40S, W20S, Sept.
1996) were much lower than expected in the presence of tar; however, equilibrium solubility
values were observed for most of the higher molecular weight compounds at these wells.
In this study, sampled groundwater showed PAH concentrations consistent with
equilibrium coal tar dissolution, indicating that equilibrium nonaqueous phase liquid (NAPL)
dissolution occurs at field sites. There is little evidence in the literature of field-scale NAPL
dissolution. Expected equilibrium groundwater concentrations of NAPL constituents have
been observed for the Bemidji crude oil spill (Eganhouse et al., 1996) and at other coal tar
sites (Backhus and Gschwend, 1994; Groher et al., 1990). Generally, however, at sites where
the presence of nonaqueous phase liquids was strongly suspected, aqueous compound
concentrations of up to only 1% of compound solubility have been reported (Jackson and
Mariner, 1995; Anderson et al., 1992).
Laboratory studies have suggested that nonaqueous phase liquids dissolve under mass
transfer-limited conditions (Powers et al., 1994; Geller and Hunt, 1993; Powers et al., 1992);
however, microscopic mass transfer limitations are likely unimportant at the macroscopic field
scale. The groundwater velocities at the monitoring well clusters at Site YYZ were high,
even under ambient conditions (0.3 - 0.6 m/d). According to laboratory studies, these pore
water velocities should result in mass transfer-limited NAPL dissolution (Geller and Hunt,
~1-~--1~--------------- -- ~.~- ..1~--
--- ~
'----- ---~--11-- ----n------- -Y- --Y---rr~.x.~ ~~_..~----
~~ I~~~-~-~--~qil
-----i ..-.-.
I-LiYI"~Y'I~""-~"~~~L-~~--LIII~--- -~-l-llllltll ---
1993; Powers et al., 1992). In addition, dissolution from coal tar may be further hindered by
the presence of skins on aged tar blobs (Luthy et al., 1993). These NAPL dissolution studies
were conducted in small columns of 5 to 15 cm dimensions (Geller and Hunt, 1993; Powers
et al., 1992). At larger flow scales, longer contact times between the groundwater and the
nonaqueous phase may allow constituent concentrations to build up to equilibrium levels.
Mathematically, equilibrium concentrations have been predicted to occur after 10 cm travel
distances at pore water velocities of 0.86 m/d (Miller et al., 1990). Equilibrium
concentrations have been observed experimentally for a 15 cm source region within a 0.75 m 3
flow system (Anderson et al., 1992). The coal tar distribution upgradient of the study area at
Site YYZ is not known, but groundwater may have been in contact with dissolving tar for up
to 300 m before it flowed past the monitoring wells, allowing equilibrium aqueous
hydrocarbon concentrations to build up at this field site.
Lower-than-equilibrium concentrations of NAPL constituents at other field sites likely
result from hydraulic dilution. Dilution would affect all constituents of a multi-component
NAPL similarly, with slight variations according to compound diffusivities. Observed
groundwater concentrations would fall below the 1:1 line on a plot versus expected
concentration by an equal amount, independent of compound solubility. Thus ratios of
compound concentrations would remain constant in the presence of a multi-component NAPL
source. At one field site with a two-component NAPL source, ratios of aqueous compound
concentrations were constant, despite being below expected equilibrium concentrations
(Jackson and Mariner, 1995). As the scale of groundwater observation points grow beyond
the scale of the NAPL distribution in the subsurface, hydraulic dilution likely becomes more
important. Equilibrium solute concentrations fall off sharply away from residual nonaqueous
phase liquid sources (Jackson and Mariner, 1995; Whelan et al., 1994; Anderson et al., 1992).
Thus, the lack of hydraulic dilution at Site YYZ suggests that all of the monitoring wells are
located in close proximity to coal tar sources and tar is widely distributed through the fill
solids at Site YYZ.
Facilitated Transport
Monitoring wells that had enhanced groundwater concentrations of PAHs, that is, in
excess of tar-water equilibrium values, showed an increasing ratio of observed-to-equilibrium
concentrations with increasing compound hydrophobicity. This trend was most clearly noted
at wells W40M and W20S (Figures 2.8 and 2.12) at which the lowest solubility compounds
deviated the greatest from their expected values in equilibrium with tar (Table 2.3). The fact
that the observed enhancements in groundwater concentrations at this site increase with
compound hydrophobicity (or particle reactivity) is suggestive of a second phase present in
the groundwater which can facilitate the transport of these coal tar PAHs.
We again address whether the enhanced load of groundwater PAHs was truly mobile
at this site or whether the observed groundwater concentrations were an artifact of our
sampling procedure. As discussed previously, groundwater turbidities and particle sizes
suggest that it is unlikely that our sampling procedure introduced non-mobile particles to our
groundwater samples and caused the elevated PAH concentrations. One potential artifact that
was not addressed that in the previous discussion was the presence of mobile nonaqueous
phase liquids.
Mobile NAPL may flow into and be retained within monitoring wells at Site YYZ and
other NAPL-contaminated sites. Care must be taken to ensure that this phase is not
inadvertently sampled in addition to the groundwater, causing higher than true contaminant
concentrations in the groundwater sample. With tar concentrations of PAHs many orders of
magnitude greater than their equilibrium aqueous concentrations, the entrainment of only a
small amount of tar (i.e., 10-100 pg/L quantities, see Table 2.3) in a groundwater sample may
greatly elevate PAH concentrations. The contribution of entrained tar to observed
concentrations becomes greater for PAHs with increasingly lower aqueous solubilities. If
droplets of tar were entrained from the monitoring wells while groundwater pumping, the
elevated polycyclic aromatic hydrocarbon concentrations should correlate with the presence of
tar in monitoring wells.
~I ----- ------- ~----I-------..
lll..-~.-.I.X~1-~P
--- l.I~---L--~-^IP~X^--_li~ ~I^_~-~.^^---- --^^-1111------ _-W I_ ..I*Y~C-I----- ~~~~~C~ I
--~~~-~-
Table 2.3. Calculated enhancements in polycyclic aromatic hydrocarbon concentrations at
wells W20S and W40M in Sept., 1996. Compound abbreviations are given in Table 2.2.
PY
BA
BaP
BP
OBSERVED GROUNDWATER CONCENTRATION ENRICHMENT
ABOVE TAR-WATER EQUILIBRIUM
W20S
1
2
2
2
W40M
3
8
14
12
7.1
PHYSICAL PROPERTIES
log Kow
5.13
5.91
6.50
Koc (mL/g)t
104 76
10555
106
Ktar (mL/g)
1068
1067
1070
E =
CgroundwateJCequilibrium
14
106
75
1069
(this thesis)
=
1 + (phase)K,
QUANTITY OF ORGANIC MATTER NECESSARY TO EXPLAIN ENRICHMENTS
10 mg/L
organic carbon
2
4
15
56
1 mg/L
tar
7
8
14
12
t log Koc = log K,,ow + 0.42 (Schwarzenbach et al., 1993).
No clear correlation was found between the extent of concentration enhancement for a
particular PAH and the presence of tar in the monitoring wells. Tar was always observed at
the bottom and on the sides of W40M and enhanced groundwater concentrations of the most
hydrophobic PAHs were always observed at this well (Figure 2.8 to 12), even in Dec. 1995
when no tar was observed on the surfaces of the retrieved packer and sampling line. In Dec.
1995, both the presence of tar and concentration enhancements were observed at W20S.
Similar depths (6") of tar were also observed at W100M at this sample date, but no
enhancement of PAH concentrations were observed at that well (Figure 2.8). At later
sampling dates, PAHs elevated above measured tar-water equilibrium concentrations were
observed at W100S and W100M (May '96) and W20S (Sept. '96), but no evidence of tar was
found on the scope lines or the retrieved packer assemblies. If droplets of tar were entrained
from the bottom of the monitoring wells during groundwater sampling, no elevation in PAH
concentrations should have been observed at these latter three wells, but elevated PAH
concentrations should have been observed at W100M in Dec. '95. Thus, entrainment of
mobile tar does not appear to explain these concentration enhancements.
The observed groundwater concentrations at this site are the first field evidence of
enhanced mobile loads of organic contaminants. The groundwater transport of PAHs at Site
YYZ is presumably facilitated relative to dissolved species. Possible means to allow
enhanced mobile loads of PAHs include the presence of inorganic particles, cosolvents, and
organic or organic-coated colloids in the groundwater. In order to investigate which of these
mechanisms is facilitating transport of PAHs at this site, we considered the theoretical
expression for the expected enhancement of PAH concentrations over equilibrium conditions
when a carrier phase is present in the groundwater:
E =
groundwater
=
1 + (phase)K
(2)
Cequlhbrium
where E is the enhancement factor, Cgroundwater (mg/L) is the groundwater concentration,
Cequilbrium
(mg/L) is the aqueous concentration in equilibrium with tar, (phase) (kg/L) is the
abundance of the carrier phase in the groundwater, and Kp (L/kg) is the equilibrium phasewater partition coefficient. Where the right-hand-side of Eq. 2 is dominated by the second
C"~-
Il_.~~~--I~~~~C--L^~~
--ll___~~_ll
~---II-----
--------- ~_~L-l^-~l~---_IIl
----- ~~III_~__
term, the enhancement factors of various PAHs should vary as their partition coefficients, KP.
Since partitioning to the carrier phase is driven by a compound's aqueous activity coefficient,
K, values for a series of organic compounds will have similar relative values for any
partitioning mechanism and so enhancement factors alone cannot be used to discern the
mechanism facilitating PAH transport. Additional supporting data is also required.
We first address whether cosolvents and surface active compounds could cause PAH
enhancements at Site YYZ. Both of these mechanisms would enhance compound solubilities
by lowering the groundwater surface tension. The surface tension of groundwater sampled at
W40M was calculated to be 70.6 ± 3 dyne/cm (n=5). Within the measurement variability,
this value did not differ from the surface tension measured for RO water. Much greater
variability between individual measurements was observed for the groundwater sample than
the RO water which suggests that there are surface active species present in the groundwater
at this well. The magnitude of this surface tension effect on the solubility of benzo(a)pyrene
was estimated with the Yalkowsky equation (Schwarzenbach et al., 1993):
slog
log C
C sat
=
N(cairRO -
Targw)
(3)
(HSA)
(3)
2.303RT
where Cm5 and C, (mg/L) are the aqueous concentrations in the presence and the absence,
respectively, of the surface active agent, N is Avagadro's number, (ca.
RO
-
ar gw)
(erg/cm2 ) is
the difference in surface tensions between the two water samples, R (erg/mol/K) is the gas
constant, T (K) is the temperature, and HSA (cm 2) is the molecule hydrophobic surface area,
here approximated as 250 A2 (Schwarzenbach et al., 1993). A difference of 1.4 erg/cm 2
(1 dyne/cm = 1 erg/cm 2) in surface tensions gives an increased benzo(a)pyrene (BaP)
concentration of 2.6 in the groundwater relative to RO water containing no surface active
species. The observed BaP enhancement factor at W40M was 14 in Sept. '96 when the
surface tension measurement was made. Thus, concentrations of polycyclic aromatic
hydrocarbons enhanced above tar-water equilibrium values were not explained by lowered
groundwater surface tensions alone.
The amount of cosolvent or surfactant required to enhance the benzo(a)pyrene
concentration by a factor of 14 above dissolved equilibrium with tar was estimated. A polar,
methanol-like cosolvent would need to be present at gram per liter quantities to explain this
enhancement (Schwarzenbach et al., 1993; Groher, 1989); however, only mg/L quantities of
organic carbon were measured in groundwater. On a carbon basis, surfactants and
biosurfactants enhance can elevate compound solubilities at much lower concentrations, but
not without a dramatic decrease in surface tension. For example, 50 mg/L of rhamnolipid
biosurfactants increased the apparent solubility of octadecane by a factor of 10 , but lowered
4
the surface tension of water from 72 dyne/cm to 30 dyne/cm (Zhang and Miller, 1992). Thus,
surface active species are not important contributors to the enhancement of groundwater PAH
concentrations at Site YYZ.
Inorganic particles are also potential mediators of hydrophobic compound transport in
groundwater. At Site YYZ, mineral surfaces of suspended particles are likely organic-carbon
coated, and thus groundwater solubilities of PAHs would be enhanced by partitioning into
these coatings. Filterable iron and aluminum were present at levels of 2 - 360 gM and 2 60 pm, respectively, in the groundwater (Table 2.1). Aluminum did not correlate with silicon,
thus clays were not predominant. Silica colloids may be also be present since silicon levels
were close to equilibrium with amorphous silica (Morel and Hering, 1993), and dissolved
levels were lower than total concentrations. At the groundwater pH, negatively charged silica
colloids would not sorb organic matter. The mineral particles with the potential to act as
carrier phases in Site YYZ groundwater were iron and aluminum oxides. Studies of humic
acid sorption to iron (Tipping, 1981) and aluminum (Davis, 1982) oxides would predict these
surfaces to be humic-coated at the iron, aluminum and non-purgeable organic carbon levels
(assuming is all capable of binding to mineral surfaces) in the groundwater at this site.
If organic-coated mineral particles are important facilitating phases for polycyclic
aromatic hydrocarbons, a correlation would be expected between the amount of iron and
aluminum oxides in the groundwater and the enhancement factors of PAH compounds.
Unfortunately, only W40M had PAH concentrations that were elevated by more than a factor
of 2 above tar-water equilibrium in Sept. 1996 when iron and aluminum measurements were
also made (Table 2.4). The lowest amount of filterable aluminum were observed at this
.. ~
.~~_..----~--1~-~-~s11--- ~~^c-~..l...
--~-- ----------- - c~c~.
-.---- - ------------ ~cr-..~.~.~~-r^l----~----
.I---~I~----~----~--~~ -~~-~~--F^ ~-~~l
monitoring well and the filterable iron was in the middle of the range observed at all of the
wells. Most of the monitoring wells had BaP enhancement factors between 0.4 and 1, but
they exhibited a wide range in filterable solids concentrations. Thus, the concentration of iron
or aluminum oxides was not a good predictor of enhanced groundwater concentrations.
Benzo(a)pyrene concentration enhancements were also plotted against turbidity
measurements (Figure 2.15) at all wells. Although turbidity is a bulk measure of suspended
solids, data was available from each of the well locations at all of the sampling dates. Energy
dispersive X-ray analysis suggested that the filtered groundwater particles may have had high
carbon contents. If these particles accounted for the bulk of the suspended solids in the
groundwater, they may form an important sorptive phase because of their carbon-rich
composition. As with the specific mineral particles, BaP concentrations exceeding tar-water
equilibrium levels were not found at wells with the highest turbidities. Even at a single well
location, W40M, which consistently had elevated PAH concentrations, benzo(a)pyrene
enhancement factors did not show a correlation with turbidity. Either particles are not
important sorptive phases for polycyclic aromatic hydrocarbons, or only a subclass of particles
quantified by the turbidity measurements were facilitating enhanced PAH concentrations.
Table 2.4. Enhancements in polycyclic aromatic hydrocarbon concentrations in Sept.,
1996. Enhancement factors were calculated with Equation 2. Compound abbreviations are
given in Table 2.2
Well ID
PY
CH
BaP
BP
W20S
0.6
0.6
0.4
0.3
W20M
1
2
2
2
W40S
2
1
1
0.5
W40M
3
8
14
12
W100S
1
0.9
1
1
W100M
0.9
0.7
0.6
0.4
W100D
0.7
0.6
0.6
0.5
--- I
ii
I
0
O
Cg
a)
a
a)
0
0
I r- M -
f
5
,
,
,
I
10
,
,
I ,
I
15
I
I
20
I
I
II
25
I
i
30
Turbidity (NTU)
Figure 2.15. Benzo(a)pyrene enhancement factors as a function of turbidity. Enhancement
factors were calculated with Equation 2.
Organic matter in the groundwater was present at high enough concentrations for
organic (or organic-coated) colloids to facilitate PAH transport at Site YYZ. Enhancements
in groundwater concentrations over solubility levels may occur from the partitioning of PAHs
to suspended organic matter (as has been shown in the laboratory (Chiou et al., 1987)) or to
organic-coated suspended mineral particles (Murphy et al., 1990). The abundance of organic
matter required to yield the observed enhancement factors for the most hydrophobic PAHs
were estimated with organic carbon-water (K.o) and tar-water (Kar) partition coefficients.
(Calculations with
Ktar
were made because partition coefficients for anthropogenically-
impacted sediments were greater than estimated with Koc partition coefficients likely because
of the presence of nonpolar petroleum hydrocarbons (Kile et al., 1995). At Site YYZ,
suspended organic matter may also be very nonpolar and exhibit tar-like partition
coefficients.) Organic carbon was present in groundwater at W40M at concentrations greater
than required for humic-type materials to enhance PAH concentrations to the extent observed
at this well (Tables 2.1 and 2.3).
About one third of the organic carbon measured in the
groundwater would have to be active as a sorbing phase at this well. Finally, only small
amounts of tar would be required to enhance PAHs to the extent observed at all wells (for
comparison, tar is about 95% by weight carbon, assuming pyrene is representative of tar
composition). From these estimates of the magnitude of partition coefficients, we can only
conclude that there is sufficient organic matter present in the groundwater to allow facilitated
transport of PAHs to occur at this site.
Further study was undertaken to investigate the nature of the groundwater colloid
phases. A series of separation methods, detailed in Chapter 3, were quantified the particle
bound and suspended organic colloid-associated PAHs. Over 65% of the pyrene mass in
excess of the dissolved mass in equilibrium with coal tar was associated with settled particles
in W40M groundwater (Chapter 3). The pyrene was likely particle-associated and not in tar
droplets because enhanced PAH concentrations were not correlated with the presence of tar in
monitoring wells. The particulate phase was not quantified for allow for estimation of the
pyrene particle-water partition coefficient with Eq. 2. Such estimation would indicate whether
the particles were more likely tar-coated or organic matter-coated. With the exception of
W100S, all monitoring wells at which elevated concentrations of PAHs have been observed
11--~1
...
~~1--^~.----.-.-----.~~- c~-rrc~-r,~ul---,
-- ~-~.~,_
_--...-^- -~----slr-1-I^
~~L-IX III~IY^~
-- ~--~yL------ll-C
ll---.,.I.~XI~P-.-._.I1UIIU-- --
have contained mobile tar at some time. The elevated groundwater PAH concentrations may
result from colloidal particles in close proximity to subsurface coal tar deposits which become
tar-coated. Mobile tar that flowed into (and out of) the monitoring wells may indicate the
proximity of subsurface tar. Poor core recoveries during monitoring well drilling prevented
quantification of the NAPL content of the subsurface solids at the monitoring well locations.
A second explanation for polycyclic aromatic hydrocarbon association with settled
particles is that the particles were coke fines. X-ray analysis of filterable solids gave no
detectable element peaks and carbon could not be detected with the beryllium window would
not detect carbon. A high fluorescence background was observed when coke particles were
placed in water (Chapter 5), suggesting that aromatic hydrocarbons were desorbed from the
coke (the emission spectra was similar to phenanthrene, the most soluble PAH), and thus
could be extracted from coke fines suspended in groundwater samples. Again, the
composition of the fill solids near the well screens is not known to verify the presence of
coke in the subsurface near W40M.
Suspended organic colloids were present in groundwater at each of the monitoring well
clusters at levels of about 5 mgc/L (Chapter 3), but it was unclear to what extent they
enhanced groundwater concentrations of polycyclic aromatic hydrocarbons. The fluorescence
quenching capacity of the groundwater was removed by acidification and centrifuge
ultrafiltration, or by the addition and precipitation of alum. Both of these groundwater
treatments suggested that humic acid-like organic matter was quenching the probe
fluorescence; however, this evidence was not supported by extractions of unaltered water
samples containing both the dissolved and colloid-associated PAH compounds (Chapter 3,
Appendix B). More thorough fluorescence measurements (e.g., probe lifetime measurements,
nonsorbing probes) are required to verify static probe quenching by association with organic
colloids.
Further study is required to determine whether it is tar, natural organic matter, or a
combination of both, that is enhancing groundwater PAH concentrations at Site YYZ. Since
the groundwater is at equilibrium with the dissolving coal tar, the suspended sorptive phase
also likely achieved equilibrium with dissolved tar constituents. Thus, the effect of a mobile
sorbing phase on transporting PAHs was readily observable as groundwater concentrations
---U
that were greater than expected aqueous equilibrium concentrations in the presence of tar.
Enhanced transport of PAHs has been suggested at another coal tar site with high
concentrations of chrysene and benz(a)anthracene (Backhus et al., 1993). The enhanced
groundwater concentrations we see for many compounds provide very strong evidence for
facilitated transport of organic contaminants at this coal tar site.
Biodegradation
Monitoring wells W20S and W40S had order of magnitude lower concentrations of
high solubility compounds (m+p- and o-xylene, naphthalene, and methylnaphthalenes),
relative to the tar-water equilibrium case (Figures 2.8 to 2.12). Over the entire sampling
period, o-xylene always showed smaller depletion factors (e.g.,
Cgroundwate/Cequiibrium
= 0.03)
than ethylbenzene (0.36) and the combined m- and p-xylene isomers (0.21) at W20S.
Similarly, at W40S, 2-methylnaphthalene was preferentially removed
(CgroundwaterCequilbrium
< 0.001) compared to the 1-methylnaphthalene isomer (0.3). Volatilization losses of these
compounds to the vadose zone would not discriminate within these isomer sets, nor would
volatilization be expected since W20S and W40S are screened 0.9 and 1.2 m, respectively,
below the water table. These compounds also do not readily undergo redox reactions in
groundwater. Therefore, there is likely a biological sink for low molecular weight aromatic
hydrocarbons in the field study area at Site YYZ.
A mass balance approach was employed to investigate the reasonableness of a
biological attenuation hypothesis. For any removal process, a mass balance of the reactants
and products can be conducted to determine whether their relative ratios support the
hypothesized removal mechanism. When biodegradation occurs, the required reactants and
expected products are as follows:
substrate
electron
reduced
reduced
compound +acceptor -> CO + electron + products
(4)
acceptor
Over time (or distance travelled), the groundwater would be depleted of substrate compounds
(if not continually replenished by a source) and electron acceptors. Carbon dioxide, reduced
electron acceptors and other partially mineralized products would build up. Differences in
_------
--r~----^- -
------x-
_l~_ll_-~-r~-^ICII-l~'~--~ILIP~L
--^^-r---------~---
~.I11I
11~
-1.YC1
~~~~-l_~~~l~--~
ICII---~~~~--
-I
II
groundwater composition may be expected between wells where biological removal of coal tar
constituents is occurring and those where no compounds are removed. For example, greater
CO 2 levels were present in W40S as compared to W40M which showed no loss of volatile
compounds.
The mass balance calculation was performed to assess whether biological removal of
naphthalene (by far, the predominant hydrocarbon degraded) could occur at Site YYZ. The
bulk of the purgeable organic compounds depleted from wells W20S and W40S (Table 2.1)
resulted from the loss of 10 mg/L of naphthalene, relative to the equilibrium tar dissolution
case. Complete mineralization of naphthalene would produce 30 mg/L of carbon dioxide or a
partial pressure of 101 atm. The observed partial pressures of carbon dioxide were much
greater with values of 10+° s and 10+06, respectively, at W20S and W40S. The presence of
elevated CO 2 levels is consistent with active biological degradation; however, similar CO 2
levels were also observed at the other wells which showed no losses of naphthalene.
(High levels of carbon dioxide in groundwater may result from the removal of other
compounds which were not monitored by our analytical methods, such as methane. Site YYZ
is still an active natural gas distribution center and leaking underground pipelines may be a
source of methane for microbial degradation. Alternatively, the source of elevated CO 2 levels
may be from the dissolution of inorganic carbonate sources in contact with the low pH
groundwater. MINEQL calculations indicated that under equilibrium conditions, the
groundwater was not saturated with respect to iron or calcium carbonate solids; however, this
does not preclude the presence of carbonates in the fill solids which may have kineticallylimited dissolution. A lime kiln was located near the field study area in 1870 to generate
lime for gas purification. Wastes from lime generation or gas purification may have been
disposed in the study area. Certainly gas purification wastes were found in the fill while
drilling R2.)
The stoichiometric requirements for complete mineralization of naphthalene were
calculated under aerobic and nitrate-, sulphate- and iron-reducing conditions (Figure 2.16). If
naphthalene removal occurred under any one of these degradation pathways, at least the
IF
Aromatic Hydrocarbon Biodegradation
substrate
+
electron
acceptor
CO,2
req'd
(mM)
--A
00
0.08 mM
naphthalene
metabolites
obs'd
(mM)
< 6x10
02
0.9
NO3
0.8
<2
4 -
2
0.5
1.4
Fe3 +
3.7
2 x 10
S0
reduced electron
acceptor
3
Figure 2.16. Stoichiometric electron acceptor requirements for the complete mineralization of naphthalene. The required
(req'd) amounts of electron acceptors are compared with observations (obs'd) from W40S groundwater in Sept. 1996.
(ND - not determined.)
------- ~~~-~---~~-~~~1---1
...^~111~--1111~~--- -~ ~1~1~11~-11~1
---1 III-- ---rrr--~.~ .~.~..-.-----..----.--.
I-~rICI---~----- ~-----L---.-~ --^r---r-r
I
required amount of electron acceptor would have needed to be present in the groundwater
prior to degradation. After naphthalene removal, the reduced electron acceptor species should
be present in the groundwater at levels at least as great as required for the electron acceptor.
Sufficient electron acceptors were present in the groundwater for biological removal of
low molecular weight tar constituents. Not enough oxygen was available for aerobic
degradation of naphthalene. The levels of dissolved iron (presumably Fe2+ in the reducing
groundwater) were not high enough at W20S and W40S for iron reduction to have resulted in
significant naphthalene mineralization. Nitrate reducers were also not significant sinks for
aromatic hydrocarbons at this site. The endpoint for nitrate reduction was taken to be
nitrogen gas. Measurements of nitrate suffered from high detection limits; however, nitrate
levels in the groundwater sampled on the eastern site boundary, the upgradient source of
groundwater to the study region, were below 0.1 mM (EA Engineering Science and
Technology Inc, 1993), insufficient for complete removal of naphthalene by nitrate reduction.
Sulphate levels did exceed the stoichiometric requirements at W20S, W40S, and W100S and
sulphate concentrations on the eastern site boundary ranged from 0.5 to 5 mM (EA
Engineering Science and Technology Inc, 1993). In addition, total sulphide levels sampled at
W20S and W40S were consistent with the reduction of 0.5 mM sulphate for naphthalene
removal. (Note that disposed purification wastes may also be a source of sulphide, although
groundwater was undersaturated with respect to iron sulphide according to MINEQL
calculations.) Therefore, the levels of electron acceptors and reduced species are consistent
with sulphate-reduction being the predominant biological removal process.
The aromatic hydrocarbon depletion pattern was also consistent with a sulphatereducing consortia present in the fill at Site YYZ. At wells W20S and W40S, groundwater
benzene concentrations were always at least as great as measured in equilibrium with coal tar.
The xylenes always had greater relative removals (i.e., lower
CgroundwateC equilbrium)
than
ethylbenzene, although groundwater concentrations of ethylbenzene were depleted by 0 to
50%. The ratio of groundwater-to-equilibrium concentrations of naphthalene were always less
than for the xylenes. Preferential removal of xylenes over ethylbenzene under sulphatereducing conditions has been observed in the field at the site of a fuel spill (Beller et al.,
1995). Naphthalene had a faster removal rate by sulphate-reducers than toluene or p-xylene,
I
I
in a field tracer study (Thierrin et al., 1995). No benzene removal was observed in either of
these studies. The degradation of monoaromatic hydrocarbons and naphthalene under
denitrifying conditions has only been studied in laboratory microcosms, but with conflicting
removal patterns (Kao and Borden, 1997). Naphthalene was not degraded under denitrifying
conditions (Flyvberg et al., 1993; Kuhn et al., 1988). When monoaromatic hydrocarbon
degradation was observed, ethylbenzene was removed in preference to xylenes (Kao and
Borden, 1997). Toluene has been degraded under both sulphate- and nitrate-reducing
conditions; however, the detection limits for toluene in Site YYZ groundwater were greater
than the low concentration measured in aqueous equilibrium with coal tar.
Sulphide concentrations at W100S were also present at stoichiometric proportions
according to the above naphthalene mineralization scheme, yet no removal of naphthalene was
observed. This observation may indicate that biological removal of xylenes and naphthalenes
was not occurring in the immediate vicinity of the monitoring wells. If biological removal
was occurring upstream from the monitoring wells, groundwater that was advected to WlOOS
may have passed another tar source, allowing build-up of tar compounds to equilibrium levels.
If groundwater advected to W20S and W40S had not subsequently contacted tar, depleted
levels of volatile compounds would still be exhibited, while high molecular weight PAHs with
no biological sink would still be at tar-water equilibrium concentrations.
Alternatively, biological removal rates of mono- and diaromatic hydrocarbons may be
slower than tar dissolution. The depleted groundwater pattern at W20S and W40S may
reflect a local residual tar source that has been depleted of these compounds. (The removal
of naphthalene (0.07 mol/mol of the original tar composition) would not produce an increase
in groundwater concentrations of higher molecular weight PAHs at these wells detectable by
our analytical methods.) If tar was pooled near W100S, a large reservoir of naphthalene may
be replenishing the groundwater. As noted, evidence of biological activity (sulphate and
sulphide) was observed at this well although all tar constituents were present at expected
aqueous solubilities in the groundwater.
Groundwater pumping may induce gradients that disrupt the biological processes
occurring under ambient flow conditions. First, the residence time of water flowing past a
microbial community may decrease, causing incomplete compound removal (i.e., lower
----
-----~--I~C~-I~---1IX~ II~
_a- l-*----^l-*
c-.-- .--------r.~--
1
depletion factors before pumping). At Site YYZ, this effect would be most greatly influenced
at the W20S cluster since pumping is estimated to at least triple the pore water velocity here.
No changes in compound depletion factors were observed at W20S. Secondly, pumping may
cause changes in flow patterns, drawing water into the area of differing composition than is
necessary to sustain biodegradation. No relevant groundwater quality parameters were
measured at the monitoring well clusters prior to Sept., 1996 to determine if water
composition had changed in response to pumping. As noted, reducing conditions and
hydrogen sulphide were noted at the shallow wells both before and during R2 pumping.
Thus, after 4.5 months time, the induced gradient had no effect on compound attenuation
within the field study area at Site YYZ.
We conclude that sufficient levels of electron acceptors and reduced species are
present in groundwater at Site YYZ for the biological removal of xylenes and naphthalenes by
sulphate reduction. Degradation of these mono- and diaromatic hydrocarbons by sulphatereducers has been demonstrated in the laboratory and the field (Ball and Reinhard, 1996;
Thierrin et al., 1995); however, no clear patterns of aromatic hydrocarbon degradation are
evident for specific microbial consortia (e.g., sulphate reducers v. nitrate reducers) when the
anaerobic biodegradation literature is taken as a whole. For example, xylenes may (Thierrin
et al., 1995) or may not (Flyvberg et al., 1993) be removed when toluene is degraded under
sulphate reducing conditions. The apparent differences in biodegradability of certain aromatic
hydrocarbons between studies may have resulted from impure cultures or other lacking
nutrients. Further work with known reactant inputs, allowing quantitative mass balance (e.g.
laboratory culture studies, field tracer tests), is needed to more conclusively demonstrate
biological attenuation at Site YYZ.
Summary of Results and Implications for Off-Site Transport and Remediation
Critical implications for the transport and remediation of aromatic hydrocarbons at
Site YYZ would have been missed without the quantification of compounds with a range of
six orders of magnitude in compound solubilities. At this site, use of a bulk measure of
groundwater contamination, such as total petroleum hydrocarbons, would have only captured
the distribution of naphthalenes, since they were the most abundant compounds, By including
I
compound specific analyses of many polycyclic aromatic hydrocarbons, the source of
groundwater contamination was found to be the equilibrium dissolution of residual coal tar.
Thus, the rate-limitation to remediation of the aromatic hydrocarbons in the groundwater at
Site YYZ is the presence of this nonaqueous phase liquid. Groundwater concentrations in the
fill solids are still in equilibrium with tar 50 years after the last tar production, and hence the
last possible on-site disposal, and will continue to be a source of contamination to the
adjacent river from many years. If the risk posed to aquatic organisms by groundwater
discharge of aromatic hydrocarbons to the river is greater than acceptable levels, it may be
abated by a hydraulic curtain. There are no other remedial technologies which will enable
source (residual tar) removal on the scale required for this site.
Analysis of individual aromatic hydrocarbon concentrations also showed two other fate
processes which will affect the flux of contaminants to the river from Site YYZ. First, the
presence of colloids in the groundwater will increase the flux of hydrophobic PAHs over the
flux calculated assuming tar equilibration with groundwater containing no colloids. The field
data gathered in this study suggested that there are temporal and spatial variations in the
presence of groundwater colloids at Site YYZ. In order to calculate the increased risk posed
by colloid-associated polycyclic aromatic hydrocarbons, groundwater concentration
measurements would need to be made at shoreline wells adjacent to the river bed seepage
face and a model of colloid transport through the river bed sediment would also be required.
Once in the river system, hydrophobic aromatic hydrocarbons will associate with particles and
settle to the sediment bed, exposing benthic dwelling and feeding organisms.
The second process occurring in the anthropogenic fill at Site YYZ is biological
degradation of mono- and diaromatic hydrocarbons. In this case, the flux of these compounds
to the river may be decreased relative to calculations assuming aqueous tar water equilibrium.
Again the spatial variability of microbial populations in the fill solids is must be knownn to
predict the full extent of biodegradation of groundwater aromatic hydrocarbons between the
MIT monitoring wells and the river. The sink for one and two-ringed aromatic hydrocarbons
in the river is volatative exchange to the atmosphere because of their high Henry's law
constants.
-
- ------ -------------------
-~-----I -~----------c--'~~--------~~~--~U
-----
References
CRC Handbook of Chemistry and Physics (1989). Boca Raton, FL, CRC Press, Inc.
Anderson, M. R.; Johnson, R. L.; Pankow, J. F. (1992). "Dissolution of dense chlorinated
solvents into ground water: 1. Dissolution from a well-defined residual source."
Ground Water 30: 250-256.
Backhus, D. A.; Gschwend, P. M. (1994). "Groundwater contamination by polycyclic
aromatic hydrocarbons: A case study of a coal-tar contaminated site." In Groundwater
Contamination and Control. U. Zoller, Ed. New York, Marcel Dekker: 315-335.
Backhus, D. A.; Ryan, J. N.; Groher, D. M.; MacFarlane, J. K.; Gschwend, P. M. (1993).
"Sampling colloids and colloid-associated contaminants in ground water." Ground
Water 31: 466-479.
Ball, H. A.; Reinhard, M. (1996). "Monoaromatic hydrocarbon transformation under anaerobic
conditions at Seal Beach, California: Laboratory studies." Environmental Toxicology
and Chemistry 15: 114-122.
Beller, H. R.; Ding, W.-H.; Reinhard, M. (1995). "Byproducts of anaerobic alkylbenzene
metabolism useful as indicators of in situ bioremediation." Environmental Science and
Technology 29: 2864-2870.
Chiou, C. T.; Kile, D. E.; Brinton, T. I.; Malcolm, R. L.; Leenheer, J. A.; MacCarthy, P.
(1987). "A comparison of water solubility enhancements of organic solutes by aquatic
humic materials and commercial humic acids." Environmental Science and Technology
21: 1231-1234.
Davis, J. A. (1982). "Adsorption of natural dissolved organic matter at the oxide/water
interface." Geochimica et Cosmochimica Acta 46: 2381-2393.
EA Engineering Science and Technology Inc (1993). Site CharacterizationReport for the
BG&E Spring Gardens Facility.
Eganhouse, R. P.; Dorsey, T. F.; Phinney, C. S.; Westcott, A. M. (1996). "Processes affecting
the fate of monoaromatic hydrocarbons in an aquifer contaminated by crude oil."
Environmental Science and Technology 30: 3304-3312.
Flyvberg, J.; Arvin, E.; Jensen, B. K.; Olsen, S. K. (1993). "Microbial degradation of phenols
and aromatic hydrocarbons in creosote-contaminated groundwater under
nitrate-reducing conditions." Journal of ContaminantHydrology 12: 133-150.
Geller, J. T.; Hunt, J. R. (1993). "Mass transfer from nonaqueous phase organic liquids in
water-saturated porous media." Water Resources Research 29: 833-845.
Groher, D.; Gschwend, P. M.; Backhus, D.; MacFarlane, J. K. (1990). "Colloids and sampling
groundwater to determine subsurface mobile loads." In Proc.: Environmental Research
Conference on Groundwater Quality and Waste Disposal. I. P. Muraka ,S. Cordle,
Eds.. Palo Alto, CA, Electric Power Research Institute: 18.1-18.10.
Groher, D. M. (1989). An Investigation of FactorsAffecting the Concentrationof Polycyclic
Aromatic Hydrocarbonsin Groundwaterat Coal Tar Waste Sites. M.S. Thesis,
Massachusetts Institute of Technology.
Harkins, W. D.; Brown, F. E. (1919). "The determination of surface tension (free surface
energy), and the weight of falling drops: The surface tension of water and benzene by
the capillary height method." Journalof the American Chemical Society 41: 499-524.
Jackson, R. E.; Mariner, P. E. (1995). "Estimating DNAPL composition and VOC dilution
from extraction well data." Ground Water 33: 407-414.
Kao, C.-M.; Borden, R. C. (1997). "Site-specific varability in BTEX biodegradation under
denitrifying conditions." Ground Water 35: 305-311.
Kile, D. E.; Chiou, C. T.; Zhou, H.; Li, H.; Xu, 0. (1995). "Partition of nonpolar organic
pollutants from water to soil and sediment organic matters." Environmental Science
and Technology 29: 1401-1406.
Kuhn, E. P.; Zeyer, J.; Eicher, P.; Schwarzenbach, R. P. (1988). "Anaerobic degradation of
alkylated benzenes in denitrifying laboratory aquifer columns." Applied and
Environmental Microbiology 54: 490-496.
Luthy, R. G.; Ramaswami, A.; Ghoshal, S.; Merkel, W. (1993). "Interfacial films in coal tar
nonaqueous-phase liquid-water systems." Environmental Science and Technology 27:
2914-2918.
_- ~-~~~~__r~---
-XI^-Irr^---l^---- -~----l--
l~*----L1
-~~1.
L---
-_-_---^-I-I~I-~I
_~~.~~.----
- ----------_II-_-_I___111____--II ~-_r__-_C~ -~-I~
MacKay, A. A.; Chin, Y.-P.; MacFarlane, J. K.; Gschwend, P. M. (1996). "Laboratory
assessment of BTEX soil flushing." Environmental Science and Technology 30:
3223-3231.
Miller, C. T.; Poirier-McNeill, M. M.; Mayer, A. S. (1990). "Dissolution of trapped
nonaqueous phase liquids: Mass transfer characteristics." Water Resources Research
26: 2783-2796.
Morel, F. M. M.; Hering, J. G. (1993). Principles and Applications ofAquatic Chemistry.
New York, John Wiley and Sons, Inc.
Murphy, E. M.; Zachara, J. M.; Smith, S. C. (1990). "Influence of mineral-bound humic
substances on the sorption of hydrophobic organic compounds." Environmental
Science and Technology 24: 1507-1516.
Powers, S. E.; Abriola, L. M.; Weber, W. J., Jr. (1992). "An experimental investigation of
nonaqueous phase liquid dissolution in saturated subsurface systems: Steady state mass
transfer rates." Water Resources Research 28: 2691-2705.
Powers, S. E.; Abriola, L. M.; Weber, W. J., Jr. (1994). "An experimental investigation of
nonaqueous phase liquid dissolution in saturated subsurface systems: Transient mass
transfer rates." Water Resources Research 30: 321-332.
Schecher, W. D.; McAvoy, D. C. (1994). "MINEQL+: A chemical equilibrium program for
personal computers." In Hallowell, ME, Environmental Research Software.
Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M. (1993). Environmental Organic
Chemistry. New York, NY, John Wiley & Sons, Inc.
Thierrin, J.; Davis, G. B.; Barber, C. (1995). "A groundwater tracer test with deuterated
compounds for monitoring in situ biodegradation and retardation of aromatic
hydrocarbons." Ground Water 33: 469-475.
Tipping, E. (1981). "The adsorption of aquatic humic substances by iron oxides." Geochimica
et Cosmochimica Acta 45: 191-199.
Whelan, M. P.; Voudrias, E. A.; Pearce, A. (1994). "DNAPL pool dissolution in saturated
porous media: Procedure development and preliminary results." Journal of
Contaminant Hydrology 15: 223-237.
C
Zhang, Y.; Miller, R. M. (1992). "Enhanced octadecane dispersion and biodegradation by a
Pseudomonas rhamnolipid surfactant (biosurfactant)." Applied and Environmental
Microbiology 58: 3276-3282.
~u ~~--~---.~-~II-- ~_l.-..---rx~
.l.-~-.111 --;---l--ls~-r~---II.,-...~.~. _ ~~~.~~.~ I-----
---- ~
_ ~IL~-~-~-.-
Chapter 3.
MECHANISMS OF GROUNDWATER SOLUBILITY ENHANCEMENTS OF
AROMATIC HYDROCARBONS AT A COAL TAR SITE
-- L~--
~--~-a~
-
Abstract
The colloid phases enhancing the mobile concentrations of polycyclic aromatic
hydrocarbons (PAHs) above tar-water solubility at a coal tar site were investigated by
applying a series of separation methods to groundwater samples. Over 65% of the enhanced
in situ pyrene mass was associated with settleable particles or tar droplets. The quenching of
pyrene fluorescence was decreased by groundwater treatments of acidification followed by
ultrafiltration and by addition of aluminum sulphate (alum), suggesting that humic acids
quenched pyrene fluorescence in unaltered groundwater. About 5 mgc/L of humic and fulvic
acids were present in groundwater at all monitoring wells sampled, but no enhancement in
PAH concentrations were detected over tar-water equilibrium solubility at these wells.
---- -c----------i---------"~c----
I~_
-u
--~^II1- -*-m~--~~-~-~~~~-~~~---
--
------------~--~,~-..~-~~lr-i -~
Il__rs~^LlsVI~'"^YYYlslLI~"~-1T~~'~X-~~~1~'"~ ~'~~
^---i~C-~
Introduction
In a recent investigation, polycyclic aromatic hydrocarbons were present in
groundwater at concentrations greater than expected for equilibrium dissolution of coal tar
(Chapter 2). Predictive calculations indicated that the only colloid phase present in sufficient
abundance to cause these solubility enhancements was organic matter. These calculations
could not discern between facilitating organic matter present as suspended macromolecules,
mineral particles with organic coatings, or tar droplets. The purpose of this study was to
investigate which, if any, of these organic colloids may be enhancing groundwater polycyclic
aromatic hydrocarbon (PAH) concentrations over dissolved compound equilibrium with tar.
A series of separation methods were applied to groundwater samples based on the
expected behaviors of suspended organic matter, organic-coated minerals, and tar droplets
under various separation strategies. These are summarized in Table 3.1. More than one type
of organic colloid may be removed by a single treatment, thus the order of application is
important. Tar and some bacteria may remain adhered to the walls of sample collection
vessels if the aqueous solution is carefully decanted. Particles with organic coatings can then
be separated from suspended organic matter by centrifuging the decanted sample. After
centrifugation, the supernatant could be treated to remove all, or some, of the humic
substances by precipitation with aluminum oxide floc, or acidification and ultrafiltration. At
each separation stage, a subsample of the two fractions could be collected for quantification
of the in situ compound concentrations in each, a direct measure of colloid-association.
Alternatively, each fraction could be tested for the capacity to quench the fluorescence of a
probe compound, an indirect measure of the potential for colloid-association.
The sequential separation steps summarized in Table 3.1 were applied to groundwater
samples from Site YYZ. Compound-colloid associations were quantified by the distributions
of in situ pyrene among fractions. Pyrene fluorescence quenching in separated fractions was
also measured. The nature of the organic colloids enhancing groundwater PAH concentrations
was hypothesized from the separation methods which yielded fractions which contained in situ
pyrene or which quenched pyrene probe fluorescence.
Table 3.1. Effect of separation methods on the removal of organic colloids from solution.
A 'Y' denotes removal by the treatment process. An 'N' denotes no effect.
Organic Colloid
Separation Treatment
Humic
Acid
Fulvic
Acid
Coat'd
part.
Tar
drop
Bacteria
Adheres to sampling apparatus
N
N
N
Y
Y
Settled by centrifugation
N
N
Y
Y
Y
Removed by ultrafiltration
Y
Y/N
Y
Y
Y
Precipitated by acidification
Y
N
N
N
N
Precipitated with aluminum oxide floc
Y
Y
N
N
N
I
_
~ ^~_I__III__I__PI___
.I~_I~Cl--III-LII(--(
1111
1_--^X-. _I-.-~ ..~-VICI I-~
^~IIIX
i.~LY-(-.I-I- IIIIII1IIIIC 1I~I1I--~I
Methods
Chemicals
Pyrene (Aldrich, Milwaukee, WI) and internal standards of p-terphenyl and
m-terphenyl (Ultra Scientific, North Kingstown, RI) were greater than 99% pure and used as
received. Methylene chloride, hexane and methanol solvents were OmniSolve (EM Science,
Gibbstown, NJ). Chromerge and aluminum sulphate were obtained from Fisher Scientific
(Fairlawn, NJ). Sodium carbonate was from Mallinckrodt (Paris, KY) and potassium
hydrogen phthalate was from Sigma (St. Louis, MO). Hydrochloric (Fisher) and phosphoric
(Mallinckrodt) acids were also used.
Sample Treatments
Groundwater samples were collected in September 1996 (W40M) and June 1997
(W20S, W20M, W100S, W100M) (see Chapter 2 for well locations). Samples were collected
in foil-wrapped BOD bottles by slow pumping methods (Chapter 2). They were stored at 4°C
for 2 weeks (June '97) or 5 months (Sept. '96) before use. Fluorescence quenching studies
were undertaken with samples from W40M and W100S. A duplicate sample from W40M
was fractionated for quantification of in situ pyrene distribution. Only alum-precipitated
organic matter was measured in the remaining samples from June '97.
Fractionated Extractions of Groundwater
The in situ pyrene was quantified in separated fractions of W40M groundwater.
Mineral solids were allowed to settle out of solution over the 5 month storage period. (All
particles greater than 0.4 pm in diameter would settle in this time, assuming a particle density
of 1.5 g/cm3 , omitting the effect of Brownian motion on particle resuspension. A diameter of
1.2 pm was calculated with a density of 1.05 g/cm3 .) The first step of the fractionation was
to gently siphon the supernatant from the BOD bottle, leaving a small, undisturbed volume of
water containing the settled solids. (The siphon tube was a piece of aluminum tubing primed
with purified water, introducing less than 8 mL to the 270 mL transferred volume.) The
small volume of water remaining in the original sample bottle was spiked with an internal
standard of p-terphenyl and extracted in the bottle with methylene chloride. This extract was
denoted the "settled solids + walls" fraction and contained pyrene associated with settled
solids and tar droplets which were adhered to the glass.
The siphoned supernatant contained the dissolved compounds plus those associated
with stable colloids. This fraction was subdivided by acidifying the groundwater to pH 1
with hydrochloric acid. The sample was allowed to stand for 3 days while acid-precipitated
material settled out. The siphoning procedure described above was repeated. The remaining
volume was spiked, extracted and referred to as the "pH 1 precipitate" fraction. The second
siphoned supernatant contained dissolved pyrene and organic colloids not precipitated under
acidic conditions. This volume of water was also spiked with p-terphenyl and extracted with
methylene chloride. This fraction was called the "pH 1 dissolved" fraction.
The fractionated methylene chloride extracts were transferred to hexane and analyzed
by capillary gas chromatography. The gas chromatograph was a Carlo Erba HRGC equipped
with a 30 m DB5-MS column (0.32 mm ID, 0.25 pm film thickness, J&W Scientific, Folsom,
CA) to which cold on-column injections were made. An injection standard of m-terphenyl
was added to the extract just prior to cold on-column injection to quantify the volume of the
extract. The temperature program began at 70 0 C with a ramp of 12oC/min to 120 0 C,
followed by a ramp of 30 C/min to 175 0 C, and a ramp of 80C/min to 300 0 C with a final hold
time of 5 min at 300 0 C. Compounds were detected by a flame ionization detector (300 0 C)
and quantified by measuring peak heights or integrating peak areas and comparing to known
external standards. Pyrene concentrations were internal standard corrected with p-terphenyl
recoveries.
Fluorescence Quenching
Pyrene fluorescence measurements were made with a Perkin Elmer LS50B
fluorometer. The excitation wavelength was 334 nm with a slit width of 4 nm and the
emission wavelength was 373 nm, also with a slit width of 4 nm. The absorbances at these
two wavelengths were measured on a Beckman DU 640 spectrophotometer to correct for the
inner filter effect (Gauthier et al., 1986). Fluorescence measurements were made with water
samples siphoned from the collection vessels to include only the colloid species most stable
over a 5 month period. Background fluorescence readings of a 3 mL water sample were
taken before pyrene addition. Four 50 jiL aliquots of a pyrene in methanol stock solution
were added to the cuvette. The cuvette was allowed to stand for 10 min after each addition
before fluorescence measurements were made.
In all cases, fluorescence response was linear
and quantified as the slope of a plot of background-corrected fluorescence versus pyrene
concentration. A duplicate sample with 3 mL of oxygen-free (< 0.3 pM, Chemettes,
Chemetrics, Calverton, VA) purified water (18 M9, Aries purification system, Vaponics,
Rockland, MA) was treated identically to quantify pyrene fluorescence in the absence of
quenchers. Because these were single point measurements at only one quencher
concentration, the linearity of a Stem Volmer plot was verified by a dilution series to
minimize coagulation artifacts by colloid concentration. Groundwater was diluted with
oxygen-free 18 MQ water. All sample manipulations were made in an argon or nitrogen
atmosphere. Between water samples, cuvettes were chromerged for 30 minutes, followed by a
rinsing sequence of 18 M2 water, methanol, methylene chloride, methanol and 18 Mf2 water.
Organic Carbon Measurements
Total organic carbon (TOC) in water samples was determined after removal of
inorganic carbon. Samples were acidified to pH 3 with phosphoric acid and bubbled with
nitrogen or argon for 10 minutes. TOC was measured by high temperature oxidation with a
Shimadzu TOC-5000 Organic Carbon analyzer externally calibrated with potassium hydrogen
phthalate standards. Triplicate measurements were made. The mean + the standard deviation
of these analyses is reported.
Removal of Organic Carbon
Acid precipitation. Solution pH was lowered to pH 1 by the addition of hydrochloric acid.
The acid precipitated material was separated from acid-stable organic carbon by centrifuge
ultrafiltration. Centricon 3 (Amicon, Beverley, MA) filter cartridges with a nominal cutoff of
3000 Daltons were used. Before use, the filters were washed with methanol, followed by
repeated washes with 18 Mn water until the TOC of the filtrate was indistinguishable from
18 Mf2 water. Acidified samples were ultrafiltered by centrifuging for 2 hours at 800 g.
Alum precipitation. Surface reactive organic carbon was removed from solution by addition
of alum (aluminum sulphate). Under mildly acidic conditions, a heavy aluminum hydroxide
floc formed which settled from solution, removing organic carbon which had sorbed to it.
Alum (10 mg) and sodium carbonate (5 mg to adjust pH) were added to 10 mL of water in a
centrifuge tube. The sample was shaken vigorously for 30 s and allowed to stand for 30 min
for floc formation. The floc, and associated organic matter, was separated from the
supernatant by centrifuging at 800 g for 15 minutes. The concentration of alum removed all
organic carbon from an 8 mgc/L solution of Aldrich humic acid.
Calculation of Partition Coefficients
Partition coefficients of colloidal materials were calculated from enhancement factors:
C
E =
Cw
1 +(OC)K
(1)
where E is the enhancement factor, CT (gg/mL) is the total compound concentration in a bulk
(dissolved plus colloid-associated) water sample, C, (pg/mL) is the dissolved concentration,
(OC) (g/mL) is the colloid concentration, and Ko, (mL/g) is the colloid-water partition
coefficient. While this equation has been written with organic carbon (OC) notation, it is also
applicable to mineral solid or tar-enhanced solubilities. In the case of fluorescence quenching,
the total compound concentration is given by the probe fluorescence of a sample from which
the colloid phase has been removed (F0) and the dissolved concentration is the probe
fluorescence in the sample (F,), assuming the colloid-associated probe is fully quenched.
.. ~~I------------- ~ - --------- I--~~-----....-- --
--
~-----
----
---- ~-P-~-- ---
~
-L-X_- .-
~~1~-11I~-1_ INNW-
Results and Discussion
Fractionated Extractions of Groundwater
Pyrene concentrations were elevated above dissolved concentrations in both of the
colloid-containing fractions of fractionated W40M groundwater (Table 3.2). The masses in
each fraction were summed to give the pyrene concentration which would have been obtained
by an extraction of the bulk water sample. This concentration (0.0036 ± 0.001 mg/L) was in
agreement with the concentration of pyrene (0.0046 ± 0.0013 mg/L) determined from an
extraction of 2 L of fresh W40M groundwater (Chapter 2).
The separation of the colloid fractions in W40M groundwater was likely not perfect
because gravitational settling and siphoning were used for separation. The concentration of
pyrene in the final "pH 1 dissolved" fraction was 0.0013 mg/L and a concentration of
0.0014 mg/L was measured in equilibrium with coal tar (Chapter 2). Thus, all colloidassociated pyrene appears to have been removed and the "dissolved" fraction truly contains
only dissolved pyrene. Because the pyrene concentration was not elevated in the "pH 1
dissolved" fraction, fulvic acids are likely not important sorbing colloid phases.
The importance of settled solids plus tar and acid-precipitated organic matter on the
enhancement of groundwater pyrene concentrations was calculated. In order not to suspend
settled particles, a volume of water containing dissolved and stable organic colloid-associated
pyrene was also extracted with the settled solids and walls. The portion of the pyrene mass
reported in Table 3.2 for this fraction which originated from the inclusion of dissolved and
organic colloid-associated pyrene in this volume was calculated to be 5%. This fraction was
calculated by dividing the product of the settled solids and walls volume (0.01 L) and the
effective pyrene concentration of the pH 1 dissolved and pH 1 precipitate ((0.3 x 10 3 +
0.28 x 10- ) mg/(0.23 + 0.039) L) by the mass of pyrene in the settled solids and wall fraction
(0.044 mg/L x 0.01 L). In the case of the "pH 1 precipitate", 18% of the pyrene mass was
actually dissolved and not colloid-associated. With these corrections, about 65%
((0.95x43)/(0.95x43 + 0.82x27) of the pyrene in excess of dissolved solubility in equilibrium
with tar was associated with the bottle walls or settled solids. The remainder of the excess
pyrene mass was associated with colloids that were stable over 5 months, but could be acidprecipitated.
Table 3.2. Distribution of pyrene in fractionated W40M groundwater. The fractionation
method is described in the Fractionated Extractions of Groundwater section. Bulk
groundwater values are the sum total from each of the separated fractions.
Pyrene
concentration
Volume
of extract
Mass of
pyrene
Mass
fraction of
total pyrene
(mg/L)
(L)
(mg)
(%)
pH 1 dissolved
0.0013
0.230
0.3 x 10-3
30
pH 1 precipitate
0.0071
0.039
0.28 x 10-3
27
Settled solids + walls
0.044
0.010
0.44 x 10-3
43
Bulk groundwater
0.0036
0.279
1.02 x 10-3
100
Separated fraction
-
--
I
I -- IIL-~_
~
1.
~
The distribution of pyrene in colloid fractions assumed perfect separation of settled
particles. It is possible that the "pH 1 precipitate" was settleable particles that had been
entrained in the transfer step. They would have settled from solution over the 3-day stand
time in response to gravity not acidification. Total organic carbon concentrations were not
measured in any of the fractions to verify colloid conservation or enhancement in any of the
fractions. If the colloid separation steps did achieve the proposed separation, the pyrene
enhancement factors calculated from these separations should agree with those obtained from
fluorescence quenching studies on groundwater samples subjected to selective organic carbon
removal by ultrafiltration and precipitation.
Fluorescence Quenching
Static quenching of the probe compound must be verified before fluorescence
quenching can be applied to quantify colloid-association. The presence of dissolved species,
such as oxygen, can quench pyrene fluorescence in solution, incorrectly suggesting colloidassociation of the pyrene. One way to differentiate between static and dynamic quenching of
a probe compound is with a Stern-Volmer plot (Figure 3.1). If the probe is statically
quenched, the fluorescence will be quenched according to Eq. 1 and a plot of F0 /F 1 versus
the quencher concentration (OC) will be linear. Such a plot for W40M groundwater was
linear (Figure 3.1); however, the concentration range of organic carbon was too narrow to
discern plot nonlinearities.
The Stern-Volmer plot can be interpreted in a second way if dynamic quenching of the
probe fluorescence was occurring. In this case, the quenched fluorescence is related to the
quencher concentration with the following equation:
F = 1 + k q(Q)
F1
where kq (M-'s
"
'))
(2)
is the quenching rate constant, t is the fluorescence lifetime of the probe,
and (Q) (M) is the quencher concentration. The dynamic quencher in the W40M groundwater
was assumed to be the low, but detectable levels of oxygen present. The quenching rate
1.4 -
1.3
1.1
Linear Regression Slopes
FO/F v. [OC]: 0.024 (mg C/L)
FO/F v. [02]: 0.67 (4M)
0.°.
0
2
4
6
8
10
12
14
16
18
Organic carbon (mg/L)
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Oxygen Concentration (pM)
Figure 3.1. Stern-Volmer plot of quenched pyrene fluorescence in W40M groundwater.
Quencher concentrations were varied by dilution with oxygen-free 18 MQ water.
~a--~-~-r-ll--
-L- ~
l~--~----- ------c~.~x----il~~-
x~---1'~^1
-- ~1-~-~ -- ~"--~--r~--~-- '-~-- C-
I
~I
constant for oxygen is estimated to be 1 x 1010 M-'s -' (Lakowicz, 1983) and the fluorescence
lifetime of pyrene is 200 x 10-9 S. With oxygen as the quenching species, the slope of the
Stern-Volmer plot was 670 000 M-'. A quenching rate constant of 3.4 x 1012 M-Is-1 was
estimated for oxygen with the pyrene fluorescence lifetime. This rate constant is faster than
the diffusion rate constant for oxygen, the fastest rate at which pyrene fluorescence could be
quenched by oxygen. Thus, it is likely that pyrene is statically quenched in W40M
groundwater and single-point fluorescence measurements are sufficient to quantify colloidassociation.
Pyrene fluorescence was quenched by humic acid-like organic colloids in W40M
groundwater. First, the fluorescence of pyrene in W40M groundwater was decreased relative
to a control of 18 MQ water (Table 3.3), indicating the presence of a colloid phase in the
groundwater. Ultrafiltration removed some of the quenching phase and the pyrene
fluorescence was slightly greater than in the unaltered groundwater, but did not approach the
fluorescence in 18 MC2 water. Thus, the colloid phase was able to pass through a nominal
3000 Dalton filter. Some organic-coated mineral particles or large suspended molecules may
have been retained on the filter surface, accounting for 1 mgc/L of organic matter. Mineral
particles were not expected to be significant because the water sample was siphoned from a
BOD bottle which had been stored for 5 months. When the groundwater was acidified to
pH 1 and then ultrafiltered, the probe fluorescence did approach the levels of fluorescence in
colloid-free water. Acidification of the water sample would cause humic acid-like molecules
to coagulate and be more efficiently removed from solution by the ultrafiltration. A change
in organic carbon concentration was observed between the unfiltered acidified groundwater
and the filtrate. The groundwater absorbance at 280 nm also decreased from 0.73
(ultrafiltered) to 0.5 (pH 1 ultrafiltered). (Water that was in equilibrium with tar had an
absorbance of 0.45 at 280 nm, and thus the groundwater absorbance after acidification and
ultrafiltration likely resulted from dissolved aromatic hydrocarbons.) The calculated molar
absorptivity of the removed organic matter was 690 M- cm '. This molar absorptivity is of the
same magnitude as reported for humic materials (Chin et al., 1991) and is much less than the
I
I
Table 3.3. Pyrene fluorescence in W40M groundwater after various treatments to remove
organic colloids.
Sample
Slope (fluorescence
v. pyrene concentration)
TOC
(mg C/L)
3.1 ± 0.1, 3.2 ± 0.1
1.9 + 0.4
Ultrafiltered 18 MKI water
3.9 - 0.3
3.1 - 0.1
Unaltered W40M
groundwater
2.2 - 0.1
19 - 1.5
2.4 - 0.1
17 - 0.02
pH 1 W40M groundwater
not measured
20 - 0.9
Ultrafiltered pH 1 W40M
groundwater
3.4 ± 0.2
13 - 0.4
18 MQ water
Ultrafiltered W40M
groundwater
100
---------I----.
~-~--- -^
-' II--~ II~~'~~P"l"lr~---^~~__C~~
~1---*-
C
molar absorptivity of polycyclic aromatic hydrocarbons (c.f, benz(a)anthracene absorptivity of
104 M'cm-1 (Schwarzenbach et al., 1993)). Thus, pyrene fluorescence appears to be quenched
by 4 to 7 mgc/L of organic colloids which are humic acid-like in nature.
The magnitude of the organic colloid-water partition coefficient and its enhancement
of pyrene solubility was calculated. The enhancement factor was calculated by Eq. 1 from
the ratio of fluorescence in the colloid-free solution to the fluorescence in the colloidcontaining solution. The values of Fo and F, were taken from the ultrafiltered pH 1
groundwater and the unaltered ultrafiltered groundwater because there was a known amount of
organic carbon removed from the later sample which accounted for the increased probe
fluorescence between the two samples. The enhancement factor of 1.4 was slightly less than
that calculated from the distribution of pyrene mass between the dissolved and organic colloid
phases of the fractionated extractions (Table 3.4).
Therefore, both the fractionated
extractions and the fluorescence quenching results were consistent with the presence of humic
acid in the groundwater which enhanced the concentrations of PAHs at Site YYZ above
dissolved levels in equilibrium with tar.
Pyrene fluorescence quenching studies with fresh W 00S groundwater produced
similar enhancement factors and partition coefficients as those observed for W40M
groundwater. First, pyrene fluorescence in unaltered groundwater was lower than in 18 MQ
water (Table 3.5). A control for quenchers removed by centrifugation was contaminated by
organic carbon (TOC 190 mgc/L). Sufficient sample remained to make duplicate measures of
the change in groundwater organic carbon concentration after centrifugation, but not to
remeasure pyrene fluorescence quenching. The organic carbon concentration in centrifuged
groundwater was not less than in the unaltered groundwater; thus, little pyrene sorbing
material was likely removed from solution solely by centrifugation. After removal of humic
substances by alum precipitation, W100S groundwater did not quench pyrene fluorescence
relative to the colloid-free 18 Mn water control. About 4 mgc/L was removed by alum
precipitation, increasing the fluorescence from 210 fluorescence units to 299 fluorescence
units. Again, the enhancement factor was calculated to be 1.4 with a corresponding partition
coefficient of 105 mL/goc (Table 3.4).
101
_-r
Table 3.4. Pyrene solubility enhancements by groundwater organic colloids. Partition
coefficients were calculated by applying Eq. 1.
Sample
Enhancement factor
W40M
fractionated
extractions
E =_(pH _Lprecipitate + pHIdiss'd)
pH 1 dissolved
= ((0.28 + 0.3)jig / 269mL) / 0.0013gtg/mL
= 1.7
W40M
fluorescence
quenching
E=
W100S
fluorescence
quenching
E = alum precipitated supernatant
bulk supernatant
= 299 / 210
= 1.4
pH 1 ultrafiltrate
dissolved + colloidal ultrafiltrate
= 3.4 / 2.4
= 1.4
102
[OC]
(mgC/L)
log Koc
4
5
4
5
._._-1_-~
1~1I__
.~XYCI
-.-lllllX~-LI
----i.-1IIIC
-II^LLIIII--^-
-~---- ----------- ---------C-.--1 -~--~1~_~-1II~----1-- ~411__1 _Il-._rm~-------^1~_ LII~
Table 3.5. Pyrene fluorescence quenching by W100S groundwater.
Sample
Pyrene fluorescence
TOC (mgC/L)
18 MQ water
270
Unaltered W100S groundwater
210
48 + 1
Centrifuged W100S
groundwater
151*
54 ± 1.5
Alum precipitated W100S
groundwater
299
44 + 1
*Likely reflects glassware contamination, rather than quenchers in the groundwater
sample.
103
I
The pyrene partition coefficients for organic colloids at Site YYZ were compared to
literature values. Colloid-water (Chin and Gschwend, 1992) and sediment-water partition
coefficients (Kile et al., 1995) for samples from contaminated marine environments were
greater than those observed for samples from pristine environments. By analogy, groundwater
colloids at this coal tar site may reflect the hydrocarbon-rich nature of the contaminant source,
being much less polar, and hence more sorptive, than groundwater colloids in pristine
aquifers. Such trends have been observed for colloids in other contaminated aquifers.
Organic matter upgradient of a crude oil spill (Hawley, 1996) and a sewage plume (Backhus
and Gschwend, 1990) did not quench the fluorescence of perylene probes, while organic
colloids within the plumes at both sites measurably quenched perylene fluorescence. No
upgradient samples were obtained for comparison of pyrene fluorescence quenching by
groundwater colloids within and outside of the contaminant plume at Site YYZ. Nevertheless,
this site is another example of a contaminated aquifer with groundwater organic matter that
has the capacity to sorb hydrophobic organic contaminants.
Correlation of Organic Carbon with Enhancement Factors
If the polycyclic aromatic hydrocarbon concentrations in Site YYZ groundwater are
enhanced by humic acids over dissolved levels in equilibrium with tar, the enhancement
factors of PAHs at various wells should correlate with the amount of acid-precipitated organic
matter. Unfortunately, there were no preserved Sept. '96 groundwater samples remaining from
other wells to test this hypothesis. The amount of alum-precipitated organic carbon was
measured for all of the wells sampled in June '97 (Table 3.6). Centrifugation alone had little
effect on changing the concentration of organic carbon in these groundwater samples, except
at W100M where about 3 mgc/L appeared to be particle associated. Alum precipitation
removed from 4 (W20M, W100M) to 7 mgc/L (W20S, W100S); however, insignificant
enhancements in PAH concentrations were observed at these wells (Appendix B). Alum
precipitation removes both humic and fulvic acids (VanBenschoten and Edzwald, 1990;
Edwards and Amirtharajah, 1985). If all of the carbon precipitated with the alum was fulvic
acid, no enhancement in PAH concentrations would be expected. (Pyrene showed no
association with fulvic acids at W40M.) An upper bound of 1 mgc/L of humic acids was
estimated to be present in the
104
------- .r.~-.
_-x_--r-r_
- ----- ---.-.- --.
~------~- ~-LI--
--~- ---II
li-l-l^l---L-~
Y1I-.- .--i~- -~TI~~
_~_~_~_~~~~_
Table 3.6. Organic carbon content of groundwater samples from June 1997.
Sample
Unaltered
Centrifuged
Alum precipitated
(mgC/L)
(mgC/L)
(mgC/L)
W20M
39.7 ± 0.7
40.2 ± 0.6
35.5 ± 0.4
W20S
24.4 ± 0.6
25 ± 0.2
19.0 ± 0.7
W100M
28 ± 0.6
24.9 ± 0.05
20.3 ± 0.5
W100S
28.7 ± 0.7
27.8 ± 0.1
21.2 ± 1.3
105
groundwater by reverse phase separation (Appendix B). Polycyclic aromatic hydrocarbons
associated with this level of humic acid would not be detectable as concentrations greater than
tar-water equilibrium by the extraction of groundwater samples.
While the preceding arguments about the nature of the alum-precipitated organic
carbon are consistent with no observable PAH concentration enhancements, they do not
explain the fluorescence quenching observed with W100S groundwater. The octanol water
partition coefficient (Kow) for benzo(a)pyrene (106, (Miller et al., 1985)) is 0.8 orders of
magnitude greater than the pyrene Kow (1052, (Miller et al., 1985)). To a first approximation,
the benzo(a)pyrene Koc for Site YYZ groundwater colloids would also be about 6 times
greater than the pyrene Koc, with an estimated value of 105 8. The corresponding enhancement
factor calculated for benzo(a)pyrene with Eq. 1 would then be 3.5 with 4 mg C/L organic
colloids present in the groundwater. According to the analytical precision of the liquid-liquid
extraction of 2 L groundwater samples (Chapter 2), an enhancement by a factor of 3.5 in the
concentration of benzo(a)pyrene above tar-equilibrium solubility levels should be readily
detectable but was not observed with bulk groundwater extractions (Appendix B). The June
'97 analyses only used 300 mL groundwater sample volumes rather than 2 L samples, as used
at the other sample dates (Chapter 2). Methylene chloride was added to these samples in the
lab, immediately prior to extraction, while 100 mL of methylene chloride was added to the
2 L samples in the field. Perhaps the solvent water contact times were not great enough to
extract colloid-associated compounds. The 2 L groundwater samples contacted this solvent
reservoir for a minimum of 48 hours before separatory funnel extraction.
An alternate explanation of the discrepancy between the fluorescence quenching results
and the liquid-liquid extractions of the June '97 samples was that pyrene fluorescence was not
quenched by association with organic matter. As noted, a dynamic quencher or static
inorganic quencher would reduce pyrene fluorescence, but would not sorb polycyclic aromatic
hydrocarbons from dissolving coal tar. Clearly, the alum treatment did remove the pyrene
quencher. At the pH of precipitation (pH 6), the aluminum hydroxide floc was positively
charged, thus a nonsorbing pyrene quencher must be negatively charged. If such a quencher
was present in the W40M groundwater, it would have also have had to precipitate at pH 1
and be removed by ultrafiltration. The only difference in the groundwater chemistry between
106
lll_~ ~-I^-~I1____^_
^"L--.----XII~~-~LI
-~-- II--------- --- --I-----.-.
1
the two wells was the presence of mM levels of sulphide and sulphate at W100S (Chapter 2).
Negatively charged polymeric sulphide species would have been removed from the
groundwater with the addition of alum, but it is unknown to what extent they would quench
pyrene fluorescence.
More study is required to verify that the pyrene fluorescence was quenched by
association with organic colloids in Site YYZ groundwater. The complex groundwater matrix
at this site may contain moieties that can quench the fluorescence of pyrene, but not sorb
polycyclic aromatic hydrocarbons desorbing from coal tar. Measuring the pyrene fluorescence
lifetime in solution or the use of other, non-sorbing fluorescent probes, in addition to pyrene,
may help to elucidate whether quenchers other than organic matter are present in the
groundwater at this site.
Conclusion
The fact that the enhancement factors and pyrene partition coefficients determined for
2 water samples by 2 different methods were so consistent with one another suggests the
presence of organic colloids in the groundwater at this site. On the basis of the fractionated
extractions of in situ pyrene, however, the major contributor of polycyclic aromatic
hydrocarbon concentration enhancements above tar-water equilibrium levels was particles
coated with organic-matter or tar, or tar droplets, and not organic colloids. About 5 mgc/L of
organic macromolecules was present at most wells, but the detection of colloid-enhanced PAH
groundwater concentrations by comparison of groundwater and tar-water equilibrium
extractions may not have been sensitive enough to detect organic-colloid enhancements.
107
References
Backhus, D. A.; Gschwend, P. M. (1990). "Fluorescent polycyclic aromatic hydrocarbons as
probes for studying the impact of colloids on pollutant transport in groundwater."
Environmental Science and Technology 24: 1214-1223.
Chin, Y.-P.; Gschwend, P. M. (1992). "Partitioning of polycyclic aromatic hydrocarbons to
marine porewater organic colloids." Environmental Science and Technology 26:
1621-1626.
Chin, Y.-P.; McNichol, A. P.; Gschwend, P. M. (1991). "Quantitation and characterization of
porewater organic colloids." In Organic Substances and Sediments in Water. R. A.
Baker, Ed. : 107-126.
Edwards, G. A.; Amirtharajah, A. (1985). "Removing color caused by humic acids." Journal
of the American Water Works Association 77(3): 50-57.
Gauthier, T. D.; Shame, E. C.; Guerin, W. F.; Seitz, W. R.; Grant, C. L. (1986).
"Fluorescence quenching method for determining equilibrium constants for polycyclic
aromatic hydrocarbons binding to dissolved humic materials." Environmental Science
and Technology 20: 1162-1166.
Hawley, C. M. (1996). A Field and Laboratory Study of the Mechanisms of Facilitated
Transport of Hydrophobic Organic Contaminants.M.S. Thesis, University of
Colorado.
Kile, D. E.; Chiou, C. T.; Zhou, H.; Li, H.; Xu, 0. (1995). "Partition of nonpolar organic
pollutants from water to soil and sediment organic matters." Environmental Science
and Technology 29: 1401-1406.
Lakowicz, J. R. (1983). Principles of Fluorescence Spectroscopy. New York, Plennum Press.
Miller, M. M.; Wasik, S. P.; Huang, G. L.; Shiu, W. Y.; Mackay, D. (1985). "Relationships
between octanol-water partition coefficient and aqueous solubility." Environmental
Science and Technology 19: 522-529.
Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M. (1993). Environmental Organic
Chemistry. New York, NY, John Wiley & Sons, Inc.
108
~~-----~--
---^-l~r-~---~+l
-I--~~
1_1__1 ...
)-~~1_~~~___
11~111~
--
VanBenschoten, J. E.; Edzwald, J. K. (1990). "Chemical aspects of coagulation using
aluminum salts- II. Coagulation of fulvic acid using alum and polyaluminum
chloride." Water Research 24: 1527-1535.
109
I
Chapter 4.
HYDRAULIC PROPERTIES OF FILL SOLIDS
110
------
--~lrr~--.-.~.~--.cla--
I
-( PC
-C~C--I- -^~ Il---
Abstract
The hydraulic conductivity of the anthropogenic fill solids in the water-bearing unit at
Site YYZ was calculated to be 100 m/d from analysis of tidally induced groundwater
fluctuations in the anthropogenic fill. A local hydraulic conductivity was calculated to be
30 m/d from particle size distributions of fill solids, suggesting that the hydraulic properties of
these fill materials do not vary over lengthscales from a few centimeters to tens of meters.
The groundwater velocity under ambient flow conditions was 0.6 to 2 m/d in the WSW
(2540) direction. The groundwater velocity induced by pumping R2 was in the NE (2950)
direction with magnitudes of 1.2, 0.6, and 0.2 m/d at the W20, W40, and W100 well clusters,
respectively.
111
II
Introduction
Knowledge of the hydraulic properties of an aquifer is required in order to model the
transport of groundwater contaminants. These hydraulic properties, such as porosity,
hydraulic conductivity and head gradients, are determined from a combination of laboratory
measurements and static or dynamic field tests. The hydraulic conductivity and storativity of
an aquifer, in particular, cannot be measured directly with field tests. Rather, changing head
levels in monitoring wells are recorded as a function of time or distance in response to aquifer
forcing. The data collected are compared to theoretical expressions for head changes in time
and space as functions of conductivity and storativity (i.e., type-curve matching). These
theoretical relationships were developed for natural, geologically-deposited aquifers. Saturated
zone solids composed of anthropogenic fill materials have the potential for differing hydraulic
characteristics than geologic media of the same scale. The extent to which the hydraulic
properties of and flowpaths in anthropogenic water-bearing units can be estimated from the
responses of geologically-deposited media is not known.
The purpose of this chapter was to determine the hydraulic properties (conductivity,
flow velocity) for the anthropogenic fill material at Site YYZ. The focus of this thesis was
on the physical-chemical sources and sinks affecting the groundwater transport of organic
contaminants, not on the physical flow field per se. Thus, the unique physical structure of the
fill solids and its effect on flowpaths in the fill region was not investigated as part of this
work. However, the possible unique characteristics of fill solids and the ways in which they
may differ from geologic media are discussed in general.
General Characteristics of Fill Solids
The primary contrast between fill solids and geologic media is the difference between
their depositional environments. The depositional environments give rise to the hydraulic
properties of these two aquifer types and determine which, if any, hydrology theories for
geologic media are applicable to fill solids. Landfilling likely occurs over a period of time,
yielding pockets of material that were formed at the same time, but are somewhat
discontinuous from those added prior and later. In contrast, the geologic processes through
which natural aquifers are formed tend to have distinct bedding planes which dominate the
112
.....
,....._II.
~I ..... i.~-
--------------- -
------~^- ly_1~~- 1 --11-1-~-~-
.I -----
groundwater flow direction. The bedding plane in natural aquifer sediments is generally
horizontal and gives rise to anisotropic hydraulic conductivities (Bouwer, 1978). Fill
materials that were deposited intermittently may be more isotropic with less preference for
horizontal-over-vertical flow in response to head gradients.
A second difference between anthropogenic water-bearing units and natural aquifers is
the scale of their areal extent. Industrial sites are likely on the scale of tens of hectares (104
m2 ) while geologic deposits and formations can cover thousands for meters in dimension.
(One anthropogenic exception is landfills which can be much larger in dimension than
industrial sites. Landfills are intentionally sited to not penetrate the saturated zone (Fetter,
1994); however, the transport of infiltrating precipitation may still be affected by the same
depositional characteristics as discussed here for saturated fill solids.) The large scale of
geologic media allows correlations of aquifer properties to be developed. For example,
natural aquifer sediments formed by glacial outwash show decreasing particle size with
distance from the terminus (Fetter, 1994). Such data can be used to determine the change in
hydraulic conductivities over distance in the aquifer and the scale over which conductivity can
be assumed relatively homogeneous (Thompson, 1994). It is unlikely that any sorting of
solids occurs as filled land is made. Geologic features in natural aquifers may also remain
constant over great distances. Again, the sporadic nature of landfilling may inhibit the
formation of such continuous features. Thus, anthropogenic water-bearing units have the
potential to be very heterogeneous, depending upon the range in particle sizes of the waste
materials that were landfilled.
The nature and arrangement of fill materials (e.g. coarse building rubble v. fine ash)
suggests the relevant hydrologic model for understanding and predicting contaminant transport
through these solids. The disparate-sized anthropogenic materials can assume two primary
forms of organization as they are deposited. First, fine-grained materials can fill in the intraand interparticle porosity of large particles. The resultant water-bearing formation is
structurally similar to an unsorted glacial till. Approaches for determining hydraulic
conductivities and flow fields in glacial tills may be applicable to such anthropogenic fill
solids. The second fill organization has a limited amount of interparticle mixing, and the
variously-sized anthropogenic materials form randomly distributed pockets of varied
113
~
permeability in the fill. In this case, the resultant water-bearing unit has structural
characteristics in common with a random dual-porosity fractured aquifer system (e.g.,
Barrenblatt model (Sen, 1995)) with flow occurring predominantly through the coarse solids,
around zones of fine-textured material. This type of fill structure may also include cases of
non-Darcian flow through anthropogenic solids with high intraparticle porosity that do not
consolidate (e.g., cinder blocks). Such flow channels would have characteristics similar to
karstic systems, although on a much smaller scale, in which groundwater flow is modelled as
pipe flow (Ford and Williams, 1989).
These are a few examples of fill characteristics which may be important when
understanding the hydraulics of anthropogenic water-bearing units. Each site with
anthropogenic fill is likely unique. The extent to which hydrologic models must be modified,
or less typical hydrologic models (e.g., dual-porosity fractures) used to interpret and predict
groundwater flow fields depends upon the size distribution of the fill solids. For example,
made land could be created by landfilling sand which would be homogeneous, in contrast to
building rubble which may have non-Darcian flow.
We now focus on Site YYZ as a specific example of an industrial site with a waterbearing unit composed of anthropogenic fill. Our goal was to calculate the groundwater
velocity under ambient and induced gradient conditions. Boring logs demonstrated the
patchiness of the fill solids at this site. Nevertheless, local hydraulic conductivities (from
particle size analysis) were of the same order of magnitude as regional values (from pump
tests) at this site.
Site YYZ Hydraulics
Conceptual Picture of Site YYZ Hydrology
A conceptual picture of the groundwater flow within the study region at Site YYZ was
developed from boring observations. The boring locations are shown on a map of the study
area (Figure 4.1). A cross section profile of the fill material depicts the patchiness of the fill
solids distribution (Figure 4.2). Further evidence of the extreme heterogeneity of this water
bearing unit was indicated by the variability of the core recoveries. B4 was a 7-m continuous
114
1996
2.46'
3.65'
1890
Shoreline
1996
Shoreline
1850 '
Shoreline
-9400 3.48'
2.85'
2.33'
River
3.73'
2.55'
-9500 -
W-7
B4
O
R2
2.50'
W-2
V
2.34'
O
*
o
MIT well cluster
Recovery wellhead (ft)
Soil boring
Monitoring well
W20
WT04 0
W40
2.34'
WT06
W-12
WT03 *
Well point
0
W100
W-8
-~~~1
-960 -1700
-1700
-1600
-1500
-1400
Figure 4.1. Map of Site YYZ detailing the field study area. Axes denote distance (ft) from an arbitrary origin.
The approximate locations of historic shorelines in this landfilled region are noted. April 10, 1996 ambient
head measurements (ft relative to mean sea level) in the recovery wells are noted at the well location.
-1300
WTO4
--
R2 -
60'
a)
o
gravel
B4
silt
sandy silt
o
20'
ash/
wood chips
wood chips
ash
coke/slag
ash
ea
silty
gravel
gravel
slag
o
gravel
slag
gravel/sand
natural
sand
natural
sand
Figure 4.2. Cross-section of fill material in the field study area at Site YYZ. Depth profiles
were constructed from qualitative observations of B4 material and drilling logs (locations are
noted in Figure 4.1). The density of the graphic fill pattern decreases with coarseness of
anthropogenic solids. Note the exaggerated vertical scale. The approximate water table depth
is also noted.
116
l~-rl~~i~^x
-------^--- ~----~-~--l~ll
.---~-~ -^--~-~- UIU r*rr*-ir~~.Cr-4 --ll~-C------_.--.---s~.----.~p---r~--a~------r~- rrc-
C-~--~_- ~
~C-l -~------
boring with a total recovery of 7 m. A duplicate boring made within 2 m of the B4 location
recovered only 0.6 m of material.
The coarser anthropogenic solids in this water-bearing unit were generally, located at
depth, above the natural sand and silt. Shallower solids near the water table were finer in
texture. The qualitative observations from the cross section profile were substantiated with
particle size analysis of solids from boring B4 (Figure 4.3). Again, greater amounts of coarse
material were found deeper in the saturated zone. These results suggest that most of the
groundwater flow occurs at the base of this water-bearing unit and an induced gradient would
primarily draw water at depth from this unconfined unit.
Hydraulic Conductivity
The hydraulics at Site YYZ under ambient and induced gradient conditions were
determined from static head measurements and tidal forcing in the study area. The parameters
obtained from these measurements were used to determine an average pore water velocity in
this region according to Darcy's Law:
v
K
dh
Kh
0 dx
(1)
where v (m/d) is the pore water velocity, Kh (m/d) is the hydraulic conductivity, 0 is the
porosity, and dh/dx (m/m) is the head gradient per unit distance. For the purpose of this
study, we were interested in a bulk measure of velocity from which the mean residence time
(order of magnitude value) of groundwater in this region could be calculated.
Hydraulic conductivities were first estimated from particle size analysis of the fill
solids from B4. Solids were obtained by split spoon sampling ahead of a hollow stem auger.
Core segments were obtained in acrylic liners with a 0.6 m, 7.5 cm diameter split spoon.
Solids were removed from each segment and wet sieved through brass screens (63, 125, 250
and 500 tIm) with reverse osmosis water containing 2 g/L sodium chloride and adjusted to pH
6 to match the ionic strength and pH of the groundwater. Sieved solids were dried at 105C
for 24 h and weighed. Sufficient amounts of fill solids were retained in the cores that porous
medium flow (Darcy's Law) was assumed to occur on a local scale. Hence, hydraulic
117
Hydraulic
Conductivity
Percent by Weight
0
2.4
20
40
60
80
-
100
J
3 m/d
3.1
4 m/d
4.3
90 m/d
5.5
30m/d
6.7
10m/d
I
E
>500 um
250-500
E
125-250
EMM 63-125
<63
E
Figure 4.3. Particle size analysis of anthropogenic fill materials from boring B4.
Hydraulic conductivities were estimated from the cumulative particle size distributions
according to the relationship developed by Bedinger, 1961. The water table was at a depth
of 2.4 m at the time of this core sampling.
118
I -II-~
-~
~
-^
~..rrrn~.r~l-x^--,,
-r~-L---.-"L_~,lr.
i. ---~-*I-I.II~LIII
s^LII---~LIIUI^_
_II1I---1_I~-~~II -iCl ~~ _yy_^~__ICyL--~~ CII~_~__q__~_TI___~__I
.___
conductivities were estimated from cumulative size distributions using the relationship
developed by Bedinger (1961).
Hydraulic conductivity values calculated from particle size analysis support the
conceptual picture that the bulk of groundwater flow occurs near the bottom of the waterbearing unit (Figure 4.3). Values range from 3 m/d at the water table to one order of
magnitude greater at depth in the water-bearing unit. This highly conductive material was a
mixture of gravel and slag. The estimated Kh values suggest that this mixture behaves as a
coarse sand (Bouwer, 1978). Assuming predominantly horizontal flow, the arithmetic average
hydraulic conductivity is 30 m/d at this point location.
The transmissivity (and hence hydraulic conductivity) of the anthropogenic fill solids
was also determined over a greater zone of influence with analysis of tidal-induced
fluctuations in groundwater heads. The river adjacent to Site YYZ is tidally influenced and is
in hydraulic communication with the anthropogenic water-bearing unit. The properties of the
groundwater aquifer attenuate the amplitude and introduce a time lag into the head
fluctuations relative to the tidal cycling (Erskine, 1991):
h = ho exp -x
tTsin(
- x
(2)
where h (m) is the groundwater head relative to the mean sea level; ho (m) is the amplitude of
the tidal oscillation; S is the aquifer storage; T (m2/d) is the aquifer transmissivity; to (d) is the
period of the tidal oscillation; x (m) is the distance from the shoreline, and t (d) is the time.
Aquifer parameters can be calculated from measures of the time lag as a function of the
distance from the shoreline:
(3)
Time lag = x
F_
119
_On
or from measures of the amplitude attenuation as a function of distance:
Tidal efficiency factor = exp -x
i~.
(4)
In 1987 tidal fluctuations were monitored in the river and in well points (W-2, W-12, W-7,
W-8) for 75 hours (Figure 4.4). Fill solid transmissivity was calculated from both the average
amplitude attenuations, and the average time lags (EA Engineering Science and Technology,
1987) observed over this time period at these well points for a tidal period of 0.5 d
(Figure 4.5). A storativity value of 0.03, representative of the fill at the depth of the well
points (EA Engineering Science and Technology, 1993), was used in the calculation. The
transmissivity was calculated to be 310 m'/d by equating the slope of the time lag plot v.
distance with Eq. 3. The slope of a log-linear plot of amplitude ratio v. distance (Eq. 4) gave
a transmissivity of 390 m2/d for this fill material. With an average saturated zone thickness
of 3 m, the hydraulic conductivity was between 100 and 130 m/d.
The local hydraulic conductivity (30 m/d) was is excellent agreement with the tidal
analysis value (100 m/d) that was averaged over the study region. This suggests that in the
study region, the hydraulic properties of these fill solids do not vary on the scale of 10s of
centimeters (core) to 10s of meters (tidal analysis). If the fill solids were significantly
fractured or exhibited preferential flow through channels, the tidal analysis would have
yielded hydraulic conductivities several orders of magnitude greater than from the particle size
analysis (Ford and Williams, 1989).
Groundwater Velocity
The ambient groundwater gradient was estimated from recovery well head
measurements made while R2 was not pumping (Figure 4.1). The recovery wells are fullscreened and fully-penetrate the fill material. Thus, heads in these wells reflect flow
throughout the depth of the fill and capture the more conductive, deeper zone. Tidal
120
~-., ~ I,.--___. ...._..,, - l.ll-P~~I^---l--_^-l_
~.~(--LI
_(~LII~-~IO-l~l~.. ...----1~----*-li^*-i___
C.tl.-O^l-~CI~ ~--~I_~~~~___IIT__~___C~r^---Lr~l~~
3.00
T
/
3
2.50
N
N-.s
w4i
4.00
'
/fo
I
-
3.00
2.00
. 2-
10
2
L
0
/it
2
/.oo
/
1.00
.1
7
\
/
\
I
Tide
.00
!So
3000.000.0
00
Time
(min)
Figure 4.4. Tidal fluctuations in the river and well points (refer locations in Figure 4.1)
(EA Engineering Science and Technology, 1987).
121
--
0.1
0.09
0.08
0.07
0.06
0.05
0.04
80
120
160
200
240
280
240
280
Distance from Shore (ft)
80
120
160
200
Distance from Shore (ft)
Figure 4.5. Analysis of tidal head fluctuations in the anthropogenic fill at Site YYZ. Well
point locations are shown in Figure 4.1. Data from EA Engineering, Science and
Technology, 1987.
122
.~-c
----~-~--~ --Yll.~i.... .-~c...~~--
------~"-"^~
"i ----~,..~-~^I -----'-~I~^11'~~--"--P---- --l~--C-^l
_l---11
11 -^-I-~_~I
-- -CI~P-- IP-~l-~~
-
fluctuations were damped in these wide diameter wells, as compared to shallow well points
which showed tidal amplitudes of 0.1 ft at 140 ft from the shore. A plane was fit through the
head measurements (SigmaPlot, Jandel Scientific). From this best fit, the head gradient was
found to be 0.006 in the WSW (2540) direction.
The gradient in the field study region is consistent with a site wide head gauging
conducted in 1987 (EA Engineering Science and Technology Inc, 1993). Groundwater heads
decreased from east to west at Site YYZ with the gradient becoming much less steep in the
highly transmissive fill material. Heads across the site were slightly higher to the north of the
study region. These head variations induce the slight southwesterly direction to the gradient
in the study area; but as a whole, groundwater generally discharges to the river to the west.
The tidal fluctuations of the river may increase the residence time of groundwater in
the study region by causing flow reversal. Near the end of the monitoring period the absolute
height of the river level became greater than the groundwater head in the study region (i.e.,
step input function) (Figure 4.4). Heads in the near shore wells (W-2, W-12) also increased
in absolute value, although to a lesser extent than the river, indicating inflow to this region
from the river. Note that heads in wells located east of the 1870 shoreline were greater than
the river level, maintaining a net westward flow of groundwater. Thus, the reversed nearshore gradient caused by a rising tidal cycle may slow groundwater efflux from the fill
material. The increased absolute head in the river, and hence inflow, to the fill material
occurred at the tidal maxima over the 75 h monitoring period. The minima for these cycles
were below the heads in the study region. Due to the backflow, a more precise description of
the net gradient in the study region should therefore include a head measure of the river level.
The ambient groundwater velocity in the study region was calculated ignoring effects
of groundwater reversal due to tidal influences. The location of the MIT monitoring wells far
from the river would minimize head variations due to tidal fluctuations, and thus flow reversal
of the groundwater would be damped relative to near shore locations. The hydraulic
conductivity of the fill material is 30 to 100 m/d, the porosity is assumed to be 0.3 and the
head gradient is 0.006. The ambient porewater velocity is thus 0.6 to 2 m/d.
The induced gradient at the MIT monitoring wells was calculated with a conservation
of mass equation. R2 was pumped at a rate of 27 L/min from Apr. 12, 1996. Assuming
123
1V
equal radial flow, this volumetric flowrate was converted to a velocity at distance, r, from the
pumping well by dividing by a cylindrical area of height saturated thickness, b, and radius, r:
v =
2ntrb6
(5)
At the W20, W40 and W100 well clusters, the induced pore water velocities are 1.2, 0.6 and
0.2 m/d, respectively. The resultant velocity vectors under induced gradient flow are shown
in Figure 4.6. The effect of the induced gradient was to increase the magnitude of the
groundwater velocity by a factor of 3 times at the W20 cluster, 2 times at the W40 cluster
and 1.3 times at the W100 cluster over a pre-pumping ambient velocity 0.6 m/d.
124
-9400
-9500
I
R2----..
I
W20
W41
V
*
-9600
-1700
MIT well cluster
Recovery well
1
-1600
-1500
-1400
Figure 4.6. Ambient (solid arrows) and induced (dotted arrows) groundwater velocities at the MIT monitoring well clusters.
-1300
References
Bedinger, M. S. (1961). Relation between median grain size and permeability in the Arkansas
River Valley, Arkansas. U.S.G.S. Prof. Paper 424-C, pp. 31-32.
Bouwer, H. (1978). GroundwaterHydrology. New York, McGraw-Hill.
EA Engineering Science and Technology (1987). Investigation and Remedial Alternatives for
the Area Near Tanks 10 and 11 and the North Tank Field at the Baltimore Gas and
Electric Company Spring Gardens Facility.
EA Engineering Science and Technology (1993). Pumping Test Conducted in the Oil
Recovery Area During March 1989.
EA Engineering Science and Technology Inc (1993). Site CharacterizationReport for the
BG&E Spring Gardens Facility.
Erskine, A. D. (1991). "The effect of tidal fluctuation on a coastal aquifer in the UK."
Ground Water 29: 556-562.
Fetter, C. W. (1994). Hydrogeology. Englewood Cliffs, NJ, Prentice Hall.
Ford, D. C.; Williams, P. W. (1989). Karst Geomorphology and Hydrology. Boston, Unwin
Hyman.
Sen, Z. (1995). Applied Hydrology for Scientists and Engineers. Boca Raton, CRC Press.
Thompson, K. D. (1994). The Stochastic Characterizationof GlacialAquifers Using Geologic
Information. PhD Thesis, Massachusetts Institute of Technology.
126
r----r .r.rrr I--rar-l-l-----l L)I-'~'
I"-~----- "---i
-----.~--r------l-^--~-~-~~-~
1 I
Chapter 5.
SORPTION OF HYDROPHOBIC COMPOUNDS TO FILL SOLIDS
127
--rab I-
---aa-l
Abstract
Industrial sites with water-bearing units composed of anthropogenic fill can contain a
variety of organic-carbon containing materials, in addition to natural organic matter. Fill
materials from a former manufactured gas plant site contained natural solids, box waste
(containing wood), wood fibres, and nonaqueous phase liquids. The total carbon content of
the fill solids (5 to 55% by weight) were several orders of magnitude greater than observed
for natural aquifer materials (less than 0.1% by weight). Solid-water partition coefficients
(Kd) for tar-containing natural solids and box waste and for coke were not accurately
predicted by total carbon measurements (fo.) and natural organic matter carbon-normalized
partition coefficients (K.o). Kd (= f,,cK)
values differed by factors of 2 to several orders of
magnitude from measured values. The overall partition coefficient for a mixture of organiccarbon containing sorbents must be calculated from the sum of the individual component
sorption:
TOT
Kd
n
=
i=1
f Ki
where i denotes the individual sorbent phases, f, is the mass fraction, and K, is the sorbentspecific partition coefficient. Partition coefficients (KdTOT) which accounted for the fraction of
individual sorbent components and component-specific partition coefficients were in good
agreement with measured values for tar-containing natural solids and box waste, but not coke
wastes. Coke sorption was an adsorptive process and was underestimated by use of activated
carbon as a model sorbent due to coke surface saturation by the sorbate. The gross
compositions of sorbent mixtures were quantified with elemental mass balance equations.
128
.,~---,
1111~-~ L--^ILIIIII---T~~--I~--I.-I-X-~--C-I~
--~L111-^_
I l-~LC -~1- .__. _--_ _~~-~II~ Y-l_~~i-I~- 1
I
Introduction
The groundwater transport of organic contaminants is governed by the tendency for a
compound to associate with the aquifer solids or to remain dissolved in the groundwater. The
solid-water partition coefficients for aquifer solids (Schwarzenbach and Westall, 1981), and
other environmental particles such as soils and sediments (Karickhoff et al., 1979), have been
correlated with the fraction of organic carbon in these sorbents. Research has since been
directed towards developing linear free energy relationships relating the organic carbonnormalized partition coefficients of these environmental solids to physical-chemical properties
of the sorbates (e.g., octanol-water partition coefficient (Karickhoff et al., 1979), solubility
(Schwarzenbach et al., 1993)). Subsequently, partition coefficients at other sites of study
could be estimated with a measurement of the carbon content of the aquifer solids and a
sorbate-specific carbon-normalized partition coefficient calculated from a linear free energy
relationship:
Kd =focKoc
(1)
where Kd (mL/g) is the partition coefficient, fo (go/gs) is the fraction organic carbon, and
Koc (mL/go) is the carbon-normalized partition coefficient. This approach for estimating Kd
values is supported by research in which little variability among carefully measured soil and
sediment KoC values was observed for samples obtained from regions throughout the world
(Kile et al., 1995). The accuracy of partition coefficient estimates made with Eq. 1 is
dependent upon how well the linear free energy relationship used to calculate Ko represents
the true organic carbon partition coefficient.
The broad applicability of Eq. 1 to predict partition coefficients at industrial sites may
be limited by the composition of anthropogenic fill solids at these sites. The partition
coefficients used to develop Eq. 1 and subsequent linear free energy relationships were
measured in systems in which the partitioning media were natural organic macromolecules, or
humic substances. Anthropogenic fill may be composed of other organic carbon-containing
media, such as wood or nonaqueous phase liquids, for which a sorbate's tendency to partition
from the aqueous phase differs from natural organic matter. Partition coefficients for these
129
-~s
solids may still be predicted with an approach similar to Eq. 1, but with modified definitions
of the correlating sorbent parameter (i.e, foc) or the linear free energy relationship.
There is evidence in the literature that some organic carbon-containing sorbents other
than natural organic matter require modified forms of Eq. 1 to predict sorbate partition
coefficients. For example, wood is a sorbent for which the fraction organic carbon must be
scaled to account for nonsorbing polymeric components (Severtson and Banerjee, 1996). The
wood partition coefficient does not correlate with the total carbon content which is relatively
constant among wood species, but it does correlate with the lignin content, and hence the
fraction of total carbon which composes lignin. Other sorbents exhibit carbon-normalized
partition coefficients that are greater than those calculated with natural organic matter free
energy relationships. One example is coal for which Ko, values are 1 to 1.5 orders of
magnitude greater than for natural organic matter-containing soils and sands (Grathwohl,
1990).
Anthropogenic fill solids at industrial sites may contain mixtures of many different
organic carbon-containing materials. In these cases, the bulk partition coefficient would be
the sum of the individual partition coefficients for each of the discrete sorbents:
fcKoc
Kd=
(2)
i =1
where i denotes each sorbent type. The summed contribution of individual sorbents was
required to estimate the partition coefficient for soil contaminated with transformer oil (Boyd
and Sun, 1990) in which the discrete sorbents were natural soil organic matter and oil. In a
second example, polycyclic aromatic hydrocarbon partitioning in marine sediments was
predicted by accounting for organic matter and soot components in the solid phase
(Gustafsson et al., 1997).
Scope of investigation
The study presented in this chapter had two objectives. The first was to characterize
the variety of organic carbon-containing materials found in a typical anthropogenic fill.
Solids were collected from a water-bearing unit composed of anthropogenic fill at a former
130
~-I
,_*...
.......
l--l~l----LI1-
~^~..
-
1MW
i.....
~c ~.--.
--.
Mp-~lw
manufactured gas plant (Site YYZ, Chapter 2). The source of the fill material was gas
production wastes and rubble from a local fire.
The second objective of this research was to determine the extent to which partition
coefficients for this anthropogenic fill could be estimated from Eq. 1 (i.e., bulk carbon
measurement and calculated organic matter Koc) or Eq. 2 (i.e. quantification of individual
sorbents with sorbent-specific partition coefficients). Naphthalene or pyrene sorption
isotherms were measured for samples of anthropogenic fill that formed lenses of discrete
sorbent types. The measured partition coefficients were compared to estimated values
calculated with Eq. 1 and Eq. 2. As part of this study, multi-element analysis was
investigated as a method to quantify the relative proportions of various sorbents in a bulk
mixture.
Methods
Chemicals
Naphthalene (J.T. Baker, Phillipsburg, NJ) and pyrene (Aldrich, Milwaukee, WI) were
used as received. Solutions were made up in reverse osmosis (RO) water (Ionics,
Bridgewater, PA). Methanol and methylene chloride solvents were 'Ultra-resi Analyzed' (J.T.
Baker). Pentane (J.T. Baker) and acetone (OmniSolv, EM Science, Gibbstown, NJ) were also
used. Mercuric chloride was from Fluka Chemie (Switzerland) and acetanalide was from
Perkin Elmer (Norwalk, CT).
Solids Collection
Anthropogenic fill material was obtained from a water-bearing unit at Site YYZ, a
former manufactured gas plant site. The water-bearing unit was constructed by filling a water
body over a period of decades with waste materials from the gas operations (Chapter 1).
Continuous cores were obtained by hollow stem auger and split spoon sampling (Chapter 2,
boring B4, Figure 2.1). Undisturbed samples were collected ahead of an auger drilled to
depth with a 0.6 m, 7.5 cm diameter split spoon. Core solids were collected in acrylic liners
which were promptly capped and subsequently stored at 40 C. To avoid contamination of core
samples, no drilling fluids were used.
131
Solids Characterization
Anthropogenic fill solids in the acrylic cores were first visually characterized, noting
the materials that formed distinct lenses of 10 to 15 cm thickness. The total carbon content
and the nonaqueous phase liquid content of the solids were quantified as a function of depth.
Fraction Organic Carbon
Bulk subsamples from 0.6 m intervals of the B4 core were dried, ground and analyzed
for carbon content by high temperature combustion (CHN 2400, Perkin Elmer, Norwalk, CT).
Carbon contents of isolated core materials (e.g., wood chips, coke) were also analyzed.
Nonaqueous Phase Liquids
The amount of nonaqueous phase liquids in the fill solids was quantified by shortexposure solvent extraction. Subsamples from 0.6 m intervals of the boring were extracted
with 1:1 pentane/acetone (v/v) in centrifuge tubes for 10 min. These short contact times were
used to remove entrapped nonaqueous phase liquids, but to minimize the removal of sorbed
constituents. The solid and solvent phases were separated by centrifugation at 840 g for
15 min. The total nonaqueous phase liquid content of the solids was estimated by two
methods (1) drying aliquots of the pentane extract to constant weight and (2) integrating the
total hydrocarbon content of the pentane extract gas chromatogram (described below).
Polycyclic Aromatic Hydrocarbons
Polycyclic aromatic hydrocarbon concentrations in the fill solids were quantified by
gas chromatographic analysis (Carlo Erba Fractovap) of the pentane extracts. Cold on-column
injections were made to a 30 m, 0.25 gm film thickness DB5 capillary column (J&W
Scientific, Folsom, CA). The temperature program began at 70'C and increased at 60 C/min
to 300 0 C. The flame ionization detector base was held at 300 0 C. Compound concentrations
were quantified with external standards. A subsample of the capillary zone oil, obtained from
free phase on the top of the 3 - 3.7 m core section, was diluted in hexane and analyzed with
this temperature program.
Sorption Isotherms
Sorbents
Anthropogenic solids were removed from the cores in which they formed 'lenses' of 10
to 15 cm thickness. Samples were removed from the center of the core. The solids
132
_--^I~-----i
~----^---~
-I^ -- --1-~1-11~-----^-~i-ll~~ -O-1L-D
i-----~- ^I~-I~~---
1
- -~ II
contacting the core liner were discarded to minimize sorption artifacts in the event that oil in
the bore hole had coated the inside of core liners retrieved from depths below the water table.
The solids of study were: (a) natural estuarine sediments; (b) box waste, and (c) coke. The
natural solids were isolated from the deepest core which had penetrated the sediments of the
former water body. These solids were primarily quartz with some residual coal tar streaking
observed in these solids.
The box wastes isolated from the site had been impregnated with tar. Box waste was
used to remove tar and impurities, such as hydrogen cyanide and hydrogen sulphide, from the
production gases. The isolated box waste had a strong coal tar odor and consisted of wood
chips with discontinuous mineral coatings. The mineral coatings had an orange-brown color
suggesting that they may have been iron oxides; however the composition of the mineral
component was not characterized.
The coke was a product of gas generation: the solid residue remaining after liberating
gases from heated coal. Large (5-7 cm) diameter porous chunks of coke were removed from
the core and crushed to facilitate isotherm measurements. The crushed coke was wet-sieved
multiple times to remove fines and the 710-1000 gm fraction used for isotherms.
Analysis
Fluorescence
Naphthalene and pyrene fluorescence measurements were made with a Perkin Elmer
LS50B Luminescence Spectrophotometer using quartz cuvettes. The excitation and emission
wavelengths were 268 nm and 322 nm, respectively, for naphthalene and 313 nm and 373 nm,
respectively, for pyrene. Excitation and emission slits were set to 5 nm. Self-absorbance was
corrected by the method of Gauthier et al. (1986) using absorbance measurements made with
a Beckman DU640 Spectrophotometer.
Turbidity interferences were checked with
absorbance measurements at 600 nm.
Gas Chromatography
Direct aqueous injections to a Carlo Erba HRGC 5160 gas chromatograph were used
to quantify naphthalene concentrations for the tar-impregnated box waste due to large amounts
of background interference from fluorescing compounds. Cold on-column injections were
made to a 19 m RTX-5 (Restek, Bellefonte, PA) capillary column with 5 gm film thickness
133
~ -_----
and 0.32 mm inner diameter. The carrier gas was hydrogen with a flowrate of 3 mL/min.
The GC was temperature programmed from 100 0 C to 250 0 C at 100 C/min. Compounds were
detected with a flame ionization detector with a base temperature of 250 0 C. Methanol
extracts were analyzed under these operating conditions, except the temperature program
began at 65C. Compound concentrations were quantified with external standards.
Tar Content
The tar contents of the natural solids and the tar-impregnated box waste were
determined following the method of Boyd and Sun (1990). The solids were Soxhlet-extracted
for 24 h with 5% methanol in methylene chloride. The extract was evaporated to dryness to
determine the mass of residue removed from the solids. An aliquot of this residue was
analyzed for carbon content.
Elemental Analysis
Elemental analysis for carbon, hydrogen and nitrogen were determined with a Perkin
Elmer 2400 CHN Elemental Analyzer by oxidation at 13000 C. Samples were dried and
ground to a fine powder before analysis. Acetanilide, individual sorbents (natural solids,
wood, coke) plus known mixtures of these sorbents were analyzed for CHNO by Galbraith
Laboratories (Knoxville, TN).
Surface Area
Surface area measurements were made by Porous Materials Inc. (Ithaca, NY) using
multi-point krypton BET isotherms.
Experimental
A fluorescence method was used to determine sorption isotherms for (a) naphthalene
to natural solids and extracted box waste and (b) pyrene to coke. Moist natural solids were
added to equilibration vessels to minimize sorption artifacts due to drying. Moisture contents
were determined by drying sample aliquots at 105 0 C for 24 h after determining isotherm
measurements. The vials were filled with reverse osmosis water and contained no headspace.
Biodegradation was prevented with 10 glg/mL mercuric chloride. For each isotherm point,
two duplicates and a positive control, containing no solids, were spiked with naphthalene from
a methanol stock solution, or pyrene from a 10% methylene chloride in methanol stock
solution. The maximum amount of organic solvent added to the vials was less than 0.7% of
134
u~-r_~_~_.-l~------ .-~---*X
.- ^.~li~-- -~------~L-~ ----L-~rr~s~Y-rn~ --i~-.-..-
^------~-II-~~~~~~~~ ~~
-i^
~-I IIIY-IIYIII^---l-l-~
r
--
I
the water volume. Negative controls, containing solids but no sorbate, were used to account
for background fluorescence leached from the solids in the course of the equilibration. In
addition, several replicate vials and controls were set up at the middle concentration to
monitor fluorescence with time and determine when sorption equilibrium was attained. When
supernatant fluorescence of the sorbent-containing vial was no longer decreasing, the isotherm
vials were centrifuged at 840 g for 30 minutes to separate the phases. Fluorescence of the
supernatant was measured.
A single point isotherm was determined for native naphthalene in the tar-containing
box waste. Five replicates and two positive controls were set up. Supernatant concentrations
were monitored over 15 d by direct aqueous injection (1 gL) gas chromatography. When
supernatant concentrations were no longer increasing tubes were centrifuged at 840 g for
30 min and analyzed.
Solid-to-water ratios were chosen to have about 50% mass sorbed at equilibrium based
on preliminary isotherm points. Solid-to-water ratios and equilibration times are summarized
in Table 5.1. Amber vials with teflon-lined solid screw caps (Supelco, Bellefonte, PA) were
used for extracted box waste and coke (15 mL) and natural solids (8 mL). Centrifuge tubes
(55 mL, Pyrex) with foil-lines solid caps containing teflon liners were used for tar-containing
box waste. All equilibration flasks were foil-wrapped or stored in boxes to minimize
photodegradation of sorbates.
Mass Balance
Mass balances were performed for several samples. The supernatant was removed and
the solids were extracted with 5 mL methanol. Methanol extracts were quantified by GC and
FID. Mass recoveries of naphthalene ranged from 50 (extracted box waste) to greater than
80% (natural solids and box waste).
Equations
Sorption isotherms were determined by plotting the solids concentration, Cs, against
the aqueous concentration, C,, at equilibrium. The solids concentration was calculated by
135
I
a
Table 5.1. Experimental conditions for sorption isotherms.
Sorbate-Sorbent Pair
Naphthalene &
Natural solids
Naphthalene & Box Waste
Solid-to-Water Ratio
Equilibration Time
(g/mL)
(d)
7.5
9
x 10-2
9
10-2
15
x
Naphthalene &
Extracted box waste
2.7
x
10-3
30
Pyrene & Coke waste
1.2
x
10'3
54
136
_-~--I-I
.~-_I
--,l~lu-^-i------asr.~-~----r-. --~-IC1-r~----~- III
-~^
LI-I X-III^ ^__I.
-.tl.---lll-~l-----IP-~IIllllllll~ I_ ~X1^II---~S~
F~----~-I
I
difference from the positive controls:
F V
C
C p
-(F, V -F
n
V)n
(3)
where F denotes a fluorescence measurement, V (mL) the aqueous phase volume, and M (g)
the mass. The subscripts pc, nc, w, and s refer to the positive control, the negative control,
sample aqueous phase and solid phase, respectively. The fluorescence response factor,
0 (FU.mL/Lg), was obtained from standard curves and used to convert to concentration
measurements. Isotherms were fit to both the linear sorption equation:
= KdC w
(4)
C =KC "
(5)
C
and the Freundlich equation:
where Kd (mL/g) is the solid-water partition coefficient, Kf (pLg' -mL"/g) is the Freundlich
coefficient, and n is the Freundlich exponent. Least squares fits were made to Eq. 4 and the
log-transformed Eq. 5 with SigmaPlot (Jandel Scientific). Isotherms were taken to be linear
where n (±a) was not significantly different than 1.
The final solids concentration for the coal tar-containing box waste was calculated by
difference from the initial solids concentration determined by gas chromatography analysis of
the Soxhlet extraction:
CS =C -
V C
(6)
where C, (Cpg/g) is the initial solids concentration solids concentration. A partition coefficient
for the tar-containing solids was calculated directly with Eq. 4 for this single-point isotherm
measurement.
137
a-
Results and Discussion
Characterization of Anthropogenic Fill Solids
The most striking characteristic of the anthropogenic fill material at Site YYZ was the
high carbon contents of these solids (Figure 5.1). Typical values in natural sand aquifers are
less than 1% (Holmen and Gschwend, 1997; Ball et al., 1990). The high carbon values at
this site resulted from the inclusion of wood chips (50% carbon by weight) and coke wastes
(80-95% carbon by weight) in the fill. Various carbon-containing solids were distributed
through the core as follows: (1) wood chips at 1.2, 3, and 5.2 m; (2) coke at 2.4 and 3.6 m;
(3) oil at 2.4, 3, and 3.6 m, and (4) coal tar at 4.2 m. Below 5.5 m the natural sands
underlying the historic river silt were encountered. The carbon contents of these sands were
low (0.001) and within the range of other natural aquifer solids.
The abundance of nonaqueous phase liquids (NAPLs) in the fill material decreased
with depth from a maxima near the water table as determined by the residue on evaporation
of pentane/acetone extracts (Figure 5.1). The high value at the water table coincided with the
occurrence of fuel oil in the capillary fringe at Site YYZ (Chapter 2), thus high residues on
evaporation would be expected at this depth. In fact, 40 mg/g of oil would correspond to
about 40% of the soil porosity, sufficient to be detected as free product in a monitoring well
(Mercer and Cohen, 1990). Oil was observed in shallow, water table monitoring wells within
the study region at Site YYZ.
Nonaqueous phase liquids were also quantified as total petroleum hydrocarbons.
Total petroleum hydrocarbons have been used as surrogate measures of NAPL concentration
at sites with fuel contamination (Huntley et al., 1994; Durnford et al., 1991). The
hydrocarbon masses in Site YYZ fill solids were calculated by integrating the total area under
the gas chromatograms of the pentane extracts. Total petroleum hydrocarbon measures of the
NAPL content of these solids were less than the residue after evaporation of the pentane
extracts, except at depths of 4.9 and 5.5 m (Figure 5.1). The quantity of total hydrocarbons
measured with this method is biased towards those compounds that can be resolved with the
chromatographic method. For example, only one third of the total compound mass of coal tar
liquid isolated from well W40M (refer Figure 2.1, Chapter 2) could be chromatographed
138
0.0 0.2 0.4 0.6 0.8 1.0
10
0
V
3
PAH Concentrations
(0 g/g)
Nonaqueous phase liquid content
(mg/g)
Fraction Organic Carbon
(g/g)
20
30
40
50
0
20 40 60 80 100
S
ll
V.
i-
i
V.
[
V
Residue on evaporation
Total Petroleum Hydrocarbons
Figure 5.1. Depth profiles of organic carbon, nonaqueous phase liquids and aromatic hydrocarbons
in the B4 boring, 1993. Note the change in units of measure between profiles. Organic carbon values
are averages of triplicate measurements, with error bars denoting the deviation in measurements.
Relative ratios of PAH concentrations (phenanthrene:pyrene:benz(a)anthracene:benzo(a)pyrene) are
noted with the solid phase PAH concentrations. The water table depth was 2.4 m at the time of boring.
under our analysis conditions, while all of the mass of isolated oil could be chromatographed.
If the pentane/acetone extracts removed a coal tar NAPL from the solids, the mass of total
hydrocarbons per gram of solids would be less than the residue on evaporation. The total
petroleum hydrocarbons as a function of depth in the fill solids did not appear to constitute a
constant fraction of the extract residue. Total hydrocarbon concentrations do not show the
same sharp decline in concentration from the water table as extract residues; however, both
the pentane extract residues and the total petroleum hydrocarbons were lowest at depths just
above the natural sediments.
For denser than water nonvolatile nonaqueous phase liquids like coal tar, the residue
on evaporation of a solvent extract will give a better estimate of the total mass of nonaqueous
phase liquids in a solid sample than a total petroleum hydrocarbon measure. Some of the
volatile compound mass will be lost during the evaporation procedure. This mass will likely
introduce less error to the final measurement of NAPL content than the unknown fraction of
the NAPL mass that can not be quantified by a chromatographic method. About 10% of the
mass of the W40M coal tar was composed of compounds more volatile than
dimethylnaphthalenes, while two thirds of the mass could not be chromatographed. Better
measures of the nonaqueous phase liquids could also be attained with a more nonpolar solvent
than used here. Acetone was added to the pentane to promote disaggregation of the fill solids
so that entrapped NAPL droplets could dissolve into the solvent. The solubility of isolated
coal tar and oil in this solvent mixture was not measured, but tar and oil were soluble in
methylene chloride. Thus, methylene chloride extracts were used to quantify the nonaqueous
phase liquid contents of solids used for sorption isotherm measurements.
The composition of the nonaqueous phase liquids (NAPLs) in the fill solids was
further investigated by comparison of the chromatographic "fingerprints" of the pentane
extracts with NAPLs (oil and coal tar) isolated from Site YYZ. The oil and tar each had
distinctive chromatographic signatures (e.g., unresolved compounds, ratios of polycyclic
aromatic hydrocarbons). The chromatograph of the isolated oil was 75% by weight
unresolved complex mixture (UCM), from which polycyclic aromatic hydrocarbon (PAH)
compounds could not be resolved. In contrast, the tar chromatograph had distinct PAH
compound peaks with little unresolved complex mixture (< 7% by weight) and were the
140
-~--~-~-, -I-~-i^I-~-...... ~._I...,
~~.~....~...I~..
--- ----~"l l~-Ys----1.. ...~.
Ll--_l~..l~
-- -------------- ~-~'-'
II
primary source of PAHs in the extracts. For comparison the total mass of compounds with a
molecular weight of 252 (e.g., benzo(a)pyrene) was 11 400 mg/L, for tar and only
700 mg/Lo in the oil. At all depths in the fill solids, distinct chromatographic peaks of
polycyclic aromatic hydrocarbons were observed, suggesting that coal tar was distributed
throughout the depth of the anthropogenic fill at Site YYZ. The concentrations of pyrene and
benzo(a)pyrene mirrored the profile of the nonaqueous phase liquid content determined by
pentane/acetone extract residues (Figure 5.1). The ratios of individual PAH compounds
(phenanthrene-to-pyrene-to-benz(a)anthracene-to-benzo(a)pyrene)
in the W40M tar were
5:2:1:1 (Tar Analysis, Chapter 2). The ratios of individual PAH concentrations
(phenanthrene:pyrene:benz(a)anthracene:benzo(a)pyrene)
at 2.4, 3.7 and 4.3 m depths differed
from the isolated coal tar suggesting that some weathering of the residual tar had occurred
(e.g., removal of phenanthrene relative to benzo(a)pyrene).
Sorption Isotherms
The high carbon contents of the fill solids and the prevalence of organic carbon in
sorbents other than natural organic matter prompted an interest in characterizing the solidwater partition coefficients of anthropogenic fill materials at Site YYZ. By comparison of the
total carbon content alone, Eq. 1 suggests that partition coefficients in this water-bearing unit
could be orders of magnitude different than for natural aquifers, overlooking the fact that
sorbent-specific Ko,, values may differ from natural organic matter carbon partition
coefficients. Sorption isotherms were measured for isolated natural solids, box waste, and
coke, as well as for solvent-extracted box waste. The experimental results were compared to
estimates of solid-water partition coefficients accounting for the nature of the organic carboncontaining sorbents. The inadequacy of using the total carbon content of the solids and a
natural organic matter-derived Ko (Eq. 1) was also demonstrated. Partition coefficients
calculated with this approach (Eq. 1) are subsequently referred to as traditional partition
coefficients (KIfad) because, until recently, no distinction has been made natural organic carbon
and other carbon-containing sorbents present in the solid phase. These experimental results
and calculations are summarized in Table 5.2 and are discussed more thoroughly in the
following sections.
141
Table 5.2. Summary of observed and estimated partition coefficients for anthropogenic fill
materials.
SorbateSorbent Pair
Method of
Calculation
Sorbent specific
Naphthalene &
SorbateSpecific
Partition
Coefficient
(mL/g)
Ko = 10293
K, = 1037
Sorbent
Mass
Fraction
(g/g)
Estimated
foc= 0.0071
ftar= 0.0011
6
Natural solids
Naphthalene &
Box waste
Naphthalene &
E x tracted b o x
Kd
Kd
(mL/g)
(mL/g)
+ 5.5
11.5
Traditional
Sorbent specific
KoC = 10293
= 1037
foc= 0.0078
6.6
ftar = 0.34
Kignm= 102 7 6
fgnin=0 . 0 3 3
20
+ 2700
2720
Traditional
Koc = 10293
foc = 0.24
200
Sorbent specific
Kiignm=102 76
fignm= 0 .0 3 3
20
Ktar
......................................................................................................................................
Meas'd
. ............
10
2400
300:
fo r C =
waste
Traditional
KoC = 10293
foc = 0.11
90
1 gg/mL
Naphthalene &
Sorbent specific
KAC= 102 44
(mL/m 2)
Ko
0 = 10293
fCOKE = 0.46
(m 2/g)
fo = 0.8
130
0:
6800
for Cw=
1 g/mL
KGAC=10 3 2
(mL/m2 )
fCOKE = 0.46
(m 2/g)
740:
for Cw=
5900:
for Cw=
0.1 ig/mL
0.1
tg/mL
Coke wastet
Traditional
Pyrene &
Coke wastet
Sorbent specific
Traditional
Ko
0 = 104 76
f0o
= 0.8
46 000
t GAC - granular activated carbon. See Coke Wastes text for description of calculation.
Single point values of Kd for these nonlinear isotherms were calculated at the specified
concentrations for comparison with linear Kd values.
142
-.^r~-^--i.~-- ---I~C ----------.~p~-r~rr-- --- -i~uu^*.ra~--
~~~X-------~---
1I1I~L-0---llllll^--~~-
~~-~~~-~-~~---~-1 'I
Natural Solids
A naphthalene sorption isotherm was measured for a bulk sample of natural solids
from Site YYZ. A bulk sample refers to the inclusion of coal tar (as evidenced by tar
streaking in the solids) as a sorbent medium, in addition to natural organic matter. The
naphthalene sorption isotherm for these solids appeared linear (Figure 5.2). The Freundlich fit
of the log-transformed data had an exponent of n = 2; however, the quality of a nonlinear fit
was not sufficiently better than a linear to justify the use of a more complex sorption
isotherm. The partition coefficient for natural solids was the slope of the linear regression
and had a value of 10 ± 1 mL/g for naphthalene.
The measured partition coefficient for natural solids was compared to the predicted
sorption of naphthalene. The total partition coefficient for these solids was hypothesized to
be the sum of the contributions of the natural organic matter and the residual tar:
Ns= f K +7K
KdS =fKoc +ftarKtw
)
(7)
where foo (go/gs) is the fraction organic matter carbon; Koc (mL/go) is the natural organic
carbon partition coefficient; fr (gtgs) is the fraction of tar, and Kw (mL/g.r) is the tar-water
partition coefficient. The fraction organic carbon was determined by difference from the
carbon content of the bulk solids (0.0078 g0o/gs) and the carbon removed by Soxhlet extraction
(0.63 goc/gres
x
0.0011 gres/gs) to be 0.0071 goc/g,. The original carbon partition coefficient was
estimated from a free energy relationship developed for polycyclic aromatic hydrocarbons
(Schwarzenbach et al., 1993) with data from Karickhoff (1981). The Ko for naphthalene was
850 mL/g (Kow = 103 .35 (Miller et al., 1985)). The fraction tar was assumed to be the residue
removed by Soxhlet extraction (0.0011 grm,/g,). Isolated tar from Site YYZ had a measured
tar-water partition coefficient of 103. 7 mL/gar. Substituting each of these values into Eq. 7
yielded a calculated naphthalene partition coefficient of 12 mL/g for these natural solids.
This is in excellent agreement with the observed value of 10 ± 1 mL/g. (The calculated KdNS
could be further tuned by adjusting the fraction of tar in Eq. 7. The carbon content of the
Soxhlet extracted residue was 0.63 gc/g,. Reported values of tar carbon contents are typically
greater than 80% (Electric Power Research Institute, 1993). This suggests that the
143
15
=L
10
5
0
0.0
0.5
1.0
1.5
Cw (Pg/mL)
Figure 5.2. Naphthalene sorption to natural solids. Error bars were calculated from
propagating analytical uncertainty in Eq. 3, and where not visible, are smaller than the data
symbols.
144
-~_I .C-XII~IX---~ILI
--- IIII^
I
---
extracted tar may have been 'diluted' by mineral fines retained in the residue. If the tar
content of the extracted residue were only 80% (0.63/0.8) of the residue mass, the predicted
Kd for these solids would decrease to 10 mL/g.)
The partition coefficient for natural solids was also calculated by ignoring the
inclusion of residual tar in these solids. The total carbon content for the natural solids was
0.0078 goc/gs and the naphthalene K0o was 850 mL/g. A traditional calculation of the partition
coefficient for natural solids gave a solid-water partition coefficient of 6.6 mL/g and
underestimated the observed partition coefficient. Partition coefficient estimates have often
been considered sufficient if they predict true solid-water partition coefficients to within a
factor of 2 (Garbarini and Lion, 1986; Karickhoff et al., 1979). By this criteria, our Kyad
would be sufficient to predict partitioning to these tar-containing natural solids. However, for
solids with greater fractions of residual nonaqueous phases, it becomes increasingly important
to account for their contribution to the bulk sorption of hydrophobic compounds. Residual
saturations of up to 0.04 gres/g, have been observed for heavy oils (Hunt et al., 1988)
(assuming residual saturation of 20%, porosity of 0.3, solids density of 2.5 g/cm 3, oil density
of 1.1 g/cm3 ), and partition coefficients for hydrocarbon liquids are almost an order of
magnitude greater than organic matter Ko values (e.g., naphthalene hexane-water partition
coefficient of 4300 mL/go, (Schwarzenbach et al., 1993), average naphthalene tar-water
partition coefficient of 9900 mL/go (Lee et al., 1992)).
Box Waste
The box waste was also a composite sorbent made of wood chips and mineral
deposits. The box waste isolated from Site YYZ also contained large amounts of tar,
characterized first by a distinctive tar odor, and later quantified by Soxhlet extraction. Only a
single-point isotherm was measured for the desorption of native naphthalene. The
naphthalene partition coefficient for this tar-impregnated box waste was 2400 mL/g.
A partition coefficient for box waste was predicted assuming naphthalene partitioning
from the wood and the tar to the aqueous phase. Organic compound partitioning to mineral
phases in aqueous media is unimportant when organic carbon contents exceed 0.001 (g0o/gs)
(Schwarzenbach and Westall, 1981). Mineral partitioning was thus assumed not to contribute
145
I
to the overall partition coefficient for box waste. The partitioning of organic sorbates to
wood is dominated by lignin partitioning. The naphthalene partition coefficient for tar-coated
box waste was calculated with the following equation:
K
=ftarKtw
'fIgnmK1
gn
(8)
where fgnmn (g/gs) is the fraction lignin in the solids and KI,,m (mL/g) is the naphthalene
lignin-water partition coefficient. The fraction of tar was again assumed to be the Soxhletextracted residue, 0.34 gres/gs, with a K, of 103 9 mL/g, for naphthalene. The fraction lignin
was calculated on a carbon basis since the box waste was not pure wood. The total carbon
content of the extracted wood was 0.11 goc/g.
Wood is typically 30% lignin by weight
(Parham and Gray, 1984) and wood (Wegner et al., 1992) and lignin (Xing et al., 1994;
Garbarini and Lion, 1986) are both about 50% carbon by weight. The equivalent fignm in
Eq. 8 was thus 0.033 go/g,. The Klgmn value was also calculated on a carbon basis from a
lignin-octanol free energy relationship (Chapter 6). KIgn for naphthalene is 10276. The
partition coefficient for tar-impregnated box waste was calculated to be 2700 mL/g and is in
agreement with the observed value of 2400 mL/g. Sorption to this box waste was dominated
by tar-water partitioning of naphthalene: the wood component partition coefficient was
estimated to be only 20 mL/g.
An estimate of the box waste partition coefficient by the traditional method
underestimated the observed value by a factor of 10. The total carbon content of the box
waste was 0.24 g0o/g, and with a naphthalene Ko, of 850 mL/g, the Krad was only 200 mL/g.
This box waste was a clear example of a nonaqueous phase-containing sorbent for which the
NAPL content must be quantified to correctly estimate sorbate partitioning.
146
~ --~II-------~I
---- - ~-~*t ~ ---F- "~~"l
r---il--I -
1
~-
I
__I_-_-_--
Solvent-Extracted Box Waste
Naphthalene sorption to box waste which had been solvent-extracted to remove the tar
component was also quantified. Again, the mineral component was assumed to be an
insignificant contributor to the overall sorption of naphthalene by extracted box waste. Thus,
the sorption isotherm for extracted box waste was hypothesized to reflect wood sorption. The
extracted box waste sorption isotherm for naphthalene was nonlinear with a Freundlich
exponent of 0.68 ± 0.02 (Figure 5.3). This nonlinearity may have resulted from an
inadvertent experimental artifact. The presumed wood-dominated extracted box waste
isotherm was measured before a study of nonpolar organic compound sorption to wood
particles was undertaken (Chapter 6). In the wood study the times for water-saturation by
wood were found to exceed the times for organic compound sorption. Care was not taken in
this investigation to ensure that the extracted box waste was water-saturated before
commencing the isotherm experiment. Although there was no difference between the aqueous
fluorescence of the sorbent-containing vessels after 9 days and 14 days, this apparent
equilibrium may have resulted from naphthalene uptake slowed by the rate of water
penetration into the dry sorbent matrix. If water had only penetrated a small portion of the
wood tissue, the effective lignin phase would have been smaller than anticipated and the local
sorbed naphthalene concentration greater than expected. A Freundlich exponent less than 1
suggests that the sorbed concentration would plateau at high aqueous concentrations. Such a
trend has been observed for sorption of stearic acid and phenol to saturated wood particles
(Stamm and Millet, 1941). Plateau-like behavior may have been artificially induced if only a
small portion of the wood tissue were water-saturated.
The sorptivity of the extracted box waste was greater than calculated, even if the
naphthalene isotherm was not at sorptive equilibrium. The calculated wood-dominated
partition coefficient for extracted box waste (Eq. 8, omitting the tar term) fell below the
observed data points (Figure 5.3). For comparison, the sorbed concentration in equilibrium
with a 1 gg/mL aqueous concentration was estimated to be 20 gg/g,. The actual sorbed
concentration had a minimum value of 300 gg/gs, a factor of 15 greater than the calculated
147
IL.
10000
1000
100
"/Calculated
Kd = 20 mL/g
10
0.01
0.1
1
10
C, (gg/mL)
Figure 5.3. Naphthalene sorption to extracted box waste. Error bars were calculated by
propagating analytical variability in Eq. 3, and where not visible, error bars are smaller than
the data symbols.
148
.;-CI-- I~------C--^l-l-.l~
ClfY--- LII~
__.~~--CII ~--L
---~--ll
ll-_l-I-1
I
value. This difference between the observed and calculated sorbed concentrations suggests
that the model assumed for sorptive uptake by extracted box waste was incorrect. Two
alternate models are (1) incomplete tar extraction and (2) structural alteration of the wood
sorbent during gas purification.
The amount of tar required to increase the sorption of the extracted box waste above
the wood-only isotherm was estimated. A single point effective partition coefficient of
300 mL/g was assumed for a naphthalene solids concentration of 300 Plg/g and an aqueous
concentration of 1 gLg/mL. Since both the tar and wood were here assumed to compose the
extracted box waste, Eq. 8 was used to estimate the remaining mass of tar. With previous
values of K,, fignm and Khgnm, the fraction of tar needed for an effective partition coefficient
of 300 mL/g was estimated to be 3% by weight of the extracted box waste. It is possible that
tar which had soaked into the porous wood structure was not effectively removed during the
short contact times of the Soxhlet extraction. Compounds take longer to diffuse out of wood
particles in nonpolar solvents that do not swell the wood tissue than wood particles in
aqueous media (Behr et al., 1953). Surficial tar would be effectively removed from the box
waste by Soxhlet extraction, but tar that had penetrated into the wood structure may not have
been effectively solubilized during the 10 minute contact times between solvent turnovers.
Once exposed to aqueous solution, the extracted box waste may have begun to swell, allowing
access of the naphthalene sorbate to internal tar deposits.
The structure of the wood matrix in the purifier boxes may have been altered during
the purification process causing changes which increased its sorptivity. The carbonnormalized sorption capacity of wood is increased when it is pyrolyzed and converted to
activated carbon (Rael et al., 1995). The temperature (40-50 0 C) and the relative humidity
(65%) for optimal operation of the purifier boxes (Gollmar, 1945) do not favor carbonization
of the wood matrix; however, fires were known to occur during the regeneration of the
mineral reagents with the infusion of air through the purifier boxes (Gollmar, 1945). Also,
the operating conditions of the purifier boxes tended toward acidic pHs from the gas
impurities (e.g., H2S, HCN). Under acidic conditions cellulose can be hydrolysed and thus
the lignin content of the wood matrix may be elevated above natural wood species. A lignin
content of 0.52 goc/g, on a carbon basis would be required to give an effective partition
149
coefficient of 300 mL/g. This value exceeds the carbon content of the extracted box waste
solids, thus enhanced lignin contents of the wood matrix do not explain the high naphthalene
sorption to extracted box waste.
Possible alterations to the wood support matrix are secondary to understanding the
groundwater transport of contaminants through box waste-rich solids at coal tar sites because
transport is likely dominated by tar-water partitioning. The order of gas passage through
multiple purifier boxes was rotated as the purification reagents became spent (Gollmar, 1945).
Thus, all matrix materials would have filtered tar from the process gas during their placement
as the first purification boxes of the sequence. Analysis of a typical box waste reported the
solids to be about 1% by weight tar and about 5% by weight organic matter fibres
(Environmental Research and Technology Inc and Koppers Company Inc, 1984). At these
proportions, the tar would dominate the partitioning between these solids because of its higher
tar-water partition coefficients, relative to lignin-water or organic matter-water partition
coefficients. At other industrial sites, however, alterations of wood structure by industrial
processes or weathering during burial may give differing sorption characteristics than
observed for kiln-dried wood.
150
----
--
U-^I----- --------c-Y
r-
-----
~~"""~I-~'-~ '
- ----
~r---
-"~--slllll
----I~-
Ir
--
I
Coke Wastes
A pyrene sorption isotherm was obtained for coke wastes isolated from Site YYZ fill
solids. Naphthalene showed no detectable sorption so a more hydrophobic compound was
used. The sorption isotherm for pyrene after 54 days of equilibration was nonlinear with a
Freundlich exponent of 1.3 ± 0.2 (Figure 5.4). A Freundlich exponent greater than 1 suggests
an increased affinity of pyrene for pyrene sorption at high concentrations, perhaps due to
interactions between sorbed molecules at high surface coverages (Parfitt and Rochester, 1983).
At the highest isotherm concentration, about 100 glg/g of pyrene was sorbed to the coke. This
sorbed concentration corresponded to about 3 x 1017 molecules per gram of coke. The crosssectional area of a pyrene molecule is 7 x 10-19 m2 (calculating the molecular radius from the
2
pyrene density of 1.3 g/cm 3 (1989)), and thus the surface coverage of the coke was 0.2 m of
2
pyrene per gram of coke. The specific surface area of coke was measured to be 0.46 m /g by
krypton sorption. The smaller size of krypton gas molecules may have allowed krypton to
penetrate coke micropores that were not accessible to pyrene molecules in aqueous solution,
and thus the coke surface may have been near saturation with respect to pyrene coverage.
The sorption of pyrene to coke was estimated from a model of pyrene sorption to
activated carbon. The microscopic structure of coke (Pierson, 1993) - randomly oriented
crystalline aromatic nuclei - is similar to that of activated carbon (Mattson and Mark, 1971),
and thus sorption sites on the coke surface may be similar to those on activated carbon. If
the sorption mechanism were the same for these two sorbents, the relative strength of pyrene
interaction with these sorbents could then be scaled according to their relative surface areas.
A partition model for coke was assumed:
Kdc
= fCO
Kcac
=
K
[Cor
ncc-
(9)
where fCOKE (m 2 /g) and fGAc (m2 /g) are the specific surface areas of the coke and granular
activated carbon (GAC), respectively; Kf (pg'-mLn/g) is the Freundlich parameter for GAC;
151
1000
100
10 -
10.001
0.1
0.01
cw (Itg/mL)
Figure 5.4. Pyrene sorption to coke waste. Error bars were calculated by propagating
analytical variability in Eq. 3, and where not visible, are smaller than the data symbols.
152
----- --C~~-~"-~l
--1 l 11 -I-I- I-~----------- I---- ~.-uL-r--
I-
~-1- --"C--
n is the Freundlich exponent for GAC, and C, (gg/mL) is the solution concentration that the
partition coefficient is evaluated at. KGAC was estimated to be 1600 mL/m 2 with values from
the high concentration pyrene isotherm determined by Walters and Luthy (1984) (fGAC =
1000 m2/g; Kf = 389 000 tg''n-mLn/g; n = 0.389). With the specific surface area of 0.46 m2 /g,
and an aqueous concentration of 0.1 jtg/mL, the pyrene partition coefficient for coke was
calculated to be 740 mL/g. This value is about an order of magnitude lower than measured
(5900 mL/g at an aqueous concentration of 0.1 tg/mL) for the coke isotherm.
The difference between the measured coke partition coefficient and the partition
coefficient estimated from pyrene sorption to activated carbon may be due to differences in
sorbent-sorbate interactions for these two solid phases. The abundance of surface sorption
sites was much lower for coke due to its smaller surface area (c.f, 1000 m2/g for 74 p~m
diameter activated carbon (Walters and Luthy, 1984)), leading to near saturated surface
coverage at the sorbed concentrations of the isotherm measurement. Less than 1% of the
activated carbon surface would have been covered at sorbed pyrene concentrations in this
range. Additional sorptive interactions, resulting from high sorbate surface coverages, may
not have been represented in the activated carbon sorption isotherm.
For comparison, the traditional partition coefficient was also calculated for pyrene
sorption to coke. The coke was 80% carbon by weight. A natural organic carbon partition
coefficient for pyrene was calculated to be 57 500 mL/goc from the linear free energy
relationship developed with data from Karickhoff (1981) (Schwarzenbach et al., 1993). The
traditional partition coefficient for coke was calculated to be 46 000 mL/g for pyrene. A
similar calculation for naphthalene yielded a value of 6800 mL/g. Both of these estimates are
at least several orders of magnitude greater than the measured uptake of these sorbates by
coke. These calculations demonstrate that, while the partition mechanism for coke is not fully
understood, a carbon-based model for predicting solid-water exchange to coke is not accurate
for this sorbent.
Sorbent Quantification
Each of the sorbents isolated from the anthropogenic fill materials at Site YYZ
required unique expressions to estimate the overall solid-water partition coefficient. In each
153
case, applying a total carbon measurement and a calculated organic carbon partition
coefficient (Eq. 1) inaccurately estimated the observed compound partitioning. However, one
benefit of applying a single Ko, value and a bulk foc measure to estimate partition coefficients
is the simplicity in making a single carbon measurement. In this way, many samples can be
quickly analyzed to characterize the spatial heterogeneity of partition coefficients within a
region of investigation. A similarly simple measurement technique is necessary to quantify
the abundance of various organic carbon-containing sorbents in a mixture.
We investigated the use of elemental analysis, including carbon, hydrogen, nitrogen
and oxygen, to separate the amounts of individual components within a mixture. The ratio of
carbon-to-oxygen content varied greatly for the sorbents investigated here, ranging from about
1 for natural organic matter to about 40 for coke (Table 5.3). Inclusion of hydrogen and
nitrogen would allow further differentiation, such as between organic matter and wood,
because of differing ratios relative to carbon (Steelink, 1985). This method was tested by
analyzing samples of individual sorbent matrices (wood, natural solids, coke) for their
elemental (CHNO) composition. Known mixtures of these sorbents were also analyzed for
their elemental composition. Mass balance equations were then written for each of the
elements, equating the total mass observed in the mixture with the unknown mass contribution
from sorbents with known compositions (Figure 5.5). The set of matrix equations was solved
for the mass fraction of each sorbent within the mixture, subject to the constraint that the sum
of the mass fractions totalled 1.
A sample calculation is shown in Figure 5.5 for a sorbent mixture with equal masses
of carbon from natural organic matter, wood, and coke. An equation for hydrogen was not
considered as one of the three elemental mass balance equations because its value was below
the detection limit for the mass of the mixtures. The elemental composition of the individual
sorbents and the mixture are also given (Figure 5.5). The mass balance constraint was
introduced into the solution by setting MNOM = 1 - MWOOD
- MCOKE.
The mass balance
equations were rewritten, substituting this expression of MNoM, to give 3 equations with 2
unknowns. The modified set of matrix equations was solved by minimizing the norm of the
error vector using Matlab. The error vector was calculated as the difference between the
154
I-L.-.YC------------------- -----------
----
- -------~---~--I
-~----~------
-II_-____I __I__
~--I.II~--~-LII
Table 5.3. Elemental composition of organic-carbon containing anthropogenic fill solids.
C
H
N
O
C/O
(g/g.)
(g/gs)
(g/gs)
(g/g)
Ratio
0.0097
0.005
0.0006
0.01
1.3
Wood
0.47
0.033
< 0.5
0.37
1.7
Coke waste
0.8
0.013
0.015
0.025
43
Solid
Natural solids
155
[C]TT = MwOOD [C]wOOD + MNOM [C]NoM
[N]TOT
=
MoKE
O [C]OKE
MWOOD [N]WOOD + MNOM [N]NOM + MCOKE [N]COKE
[O]ToT = MWOOD [O]wooD + MNOM [O]NOM + MCOKE [O]coKE
PERCENT ELEMENTAL COMPOSITION
Element
Total
Wood
Natural
Organic
Matter
C
2.77
46.64
0.97
79.86
N
0.13
< 0.5'
0.061
1.49
0
2.05
37
1.03
2.47
Coke
FRACTIONAL COMPOSITION
Solid Matrix
Known Composition
Estimated By Linear
Optimization
Wood
0.02
0.0280
Natural Organic Matter
0.96
0.9654
Coke
0.02
0.0066
Figure 5.5. Sample elemental mass balance calculation to determine the fractional
composition of a sorbent mixture.
156
------------
.~
L~- l~-- 3-^-.C^---rr~-^*~lLtl^L.CI
I
I~IY1I-
~--lyL--
II-
__ ~ilp~i~-^-^--Now"
right hand side and the left hand side of the modified versions of the equations shown in
Figure 5.5.
The mass fractional composition of the mixture calculated with this optimization
method was similar to the known composition of the mixture (Figure 5.5). The estimated
wood component was 50% greater than the actual value and the coke was underestimated by
a factor of 4, in comparison to the true mass fraction. When these fractional compositions
were substituted back into the right hand side of the mass balance equations in Figure 5.5 to
solve for the bulk elemental weight percentages, the nitrogen value was lower by 50% while
the carbon and oxygen values were the same as the values measured in the experimental
mixture, within measurement error. This check suggests that the nitrogen were the least well
known of the three elements in the elemental analysis and may result from the difficulty in
quantifying the low percentages of nitrogen in these sorbents.
The error introduced into a calculation of the overall solid-water partition coefficient
by the difference between the true and estimated fractional compositions depends upon the
relative magnitudes of the sorbent specific partition coefficients. The sorbent mixtures tested
for this elemental mass balance method were found to have less than a factor of 2 difference
between the overall partition coefficients calculated with the estimated fractional compositions
and the partition coefficients calculated with the known fractional compositions (Table 5.4).
For other sorbent mixtures, the imprecision of the elemental mass balance method may give
rise to poor estimates of the overall Kd of the mixture.
Conclusion
The sorption isotherm study demonstrated the importance of characterizing the various
organic carbon-containing materials in order to accurately predict sorption to, and hence,
groundwater transport through, anthropogenic fill materials. For example, coke wastes
composed a large fraction of the subsurface solids at Site YYZ, accounting for a total carbon
content of 35 to 75% by weight at a depth of 3.6 m; however, no detectable sorption of
naphthalene to coke was able to be measured. Thus, naphthalene retardation would be
overestimated by several orders of magnitude if this sorbent was not correctly identified and
an appropriate sorption isotherm applied. Elemental analysis can be used to determine the
157
Table 5.4. Evaluation of elemental mass balance method for determining fractional
composition of sorbent mixtures. The partition coefficients for naphthalene were calculated
with the following equation using parameter values from Table 5.2:
Kd = fwooDfignmKlgnm + fNOMfocKoc + fCOKEKCOKE
Estimated
Known Fractional
Kd from
Fractional
Composition
Estimated
Known
Composition
Composition
0.0098
0.9902
0
17
10
fNOM - 0.9868
fCOKE - 0.0062
0
0.9880
0.0124
10
9
0.0280
fNOM - 0.9654
fCOKE - 0.0066
0.0183
0.9671
0.0145
13
12
Composition
fwoOD - 0.0480
fNOM - 0.9772
fCOKE - -0.0201
fWOOD
fWOOD
-
-
0.0070
158
Kd from
~~~~
..~~ ~~..,i....~~.~x~~,.
_IPII~.~--l-XI~YII~-ILUCI
I(~
-~--- ~-~I-~--
^---_
-^-1~~1__1_
1
1~1_~~__ _~_~_
gross composition of solid matrix mixtures. Refinements of this method, such as the
inclusion of selective digestion steps, may be necessary to improve the accuracy of solidwater partition coefficient estimates for mixtures containing small mass fractions of materials
with large sorbent-specific partition coefficients.
159
I
References
CRC Handbook of Chemistry and Physics (1989). Boca Raton, FL, CRC Press, Inc.
Ball, W. P.; Buehlar, C.; Harmon, T. C.; Mackay, D. M.; Roberts, P. V. (1990).
"Characterization of a sandy aquifer material at the grain scale." Journal of
ContaminantHydrology 5: 253-295.
Behr, E. A.; Briggs, D. R.; Kaufert, F. H. (1953). "Diffusion of dissolved materials through
wood." Journal of Physical Chemistry 57: 476-480.
Boyd, S. A.; Sun, S. (1990). "Residual petroleum and polychlorinated oils as sorptive phases
for organic contaminants in soils." EnvironmentalScience and Technology 24:
142-144.
Durnford, D.; Brookman, J.; Billica, J.; Milligan, J. (1991). "LNAPL distribution in a
cohesionless soil: A field investigation and cryogenic sampler." Ground Water
Monitoring Review 11(Summer): 115-122.
Electric Power Research Institute (1993). Chemical and Physical Characteristicsof Tar
Samples From Selected Manufactured Gas Plant (MGP)Sites. EPRI TR-102184s.
Environmental Research and Technology Inc; Koppers Company Inc (1984). Handbook on
Manufactured Gas Plant Sites.
Garbarini, D. R.; Lion, L. W. (1986). "Influence of the nature of soil organics on the sorption
of toluene and trichloroethylene." Environmental Science and Technology 20:
1263-1269.
Gauthier, T. D.; Shame, E. C.; Guerin, W. F.; Seitz, W. R.; Grant, C. L. (1986).
"Fluorescence quenching method for determining equilibrium constants for polycyclic
aromatic hydrocarbons binding to dissolved humic materials." Environmental Science
and Technology 20: 1162-1166.
Gollmar, H. A. (1945). "Removal of sulfur compounds from coal gas." In Chemistry of coal
utilization. H. H. Lowry, Ed. New York, John Wiley & Sons, Inc. II: 947-1007.
Grathwohl, P. (1990). "Influence of organic matter form soils and sediments from various
origins on the sorption of some chlorinated aliphatic hydrocarbons: Implications on Koc
correlations." Environmental Science and Technology 24: 1687-1693.
160
~*--------------- ------
---------~i
----
- "-"~
--~~I------~II
~~~~-~-rs~'---~----------
- II-~-~------------
Gustafsson, O.; Haghseta, F.; Chan, C.; MacFarlane, J.; Gschwend, P. M. (1997).
"Quantification of the dilute sedimentary soot phase: Implications for PAH speciation
and bioavailability." Environmental Science and Technology 31: 203-209.
Holmen, B. A.; Gschwend, P. M. (1997). "Estimating sorption rates of hydrophobic organic
compounds in iron oxide- and aluminosilicate clay-coated aquifer sands."
Environmental Science and Technology 31: 105-113.
Hunt, J. R.; Sitar, N.; Udell, K. S. (1988). "Nonaqueous phase liquid transport and cleanup: 1.
Analysis of mechanisms." Water Resources Research 24: 1247-1258.
Huntley, D.; Hawk, R. N.; Corley, H. P. (1994). "Nonaqueous phase hydrocarbon in a
fine-grained sandstone: 1. Comparison between measured and predicted saturations and
mobility." Ground Water 32: 626-634.
Karickhoff, S. W. (1981). "Semi-empirical estimation of sorption of hydrophobic pollutants on
natural sediments and soils." Chemosphere 10: 833-846.
Karickhoff, S. W.; Brown, D. S.; Scott, T. A. (1979). "Sorption of hydrophobic pollutants on
natural sediments." Water Research 13: 241-248.
Kile, D. E.; Chiou, C. T.; Zhou, H.; Li, H.; Xu, 0. (1995). "Partition of nonpolar organic
pollutants from water to soil and sediment organic matters." Environmental Science
and Technology 29: 1401-1406.
Lee, L. S.; Rao, P. S. C.; Okuda, I. (1992). "Equilibrium partitioning of polycyclic aromatic
hydrocarbons from coal tar into water." Environmental Science and Technology 26:
2110-2115.
Mattson, J. S.; Mark, H. B. (1971). Activated Carbon. New York, Marcel Dekker, Inc.
Mercer, J. W.; Cohen, R. M. (1990). "A review of immiscible fluids in the subsurface:
Properties, models, characterization and remediation." Journal of Contaminant
Hydrology 6: 107-163.
Miller, M. M.; Wasik, S. P.; Huang, G. L.; Shiu, W. Y.; Mackay, D. (1985). "Relationships
between octanol-water partition coefficient and aqueous solubility." Environmental
Science and Technology 19: 522-529.
161
---- ~p-
Parfitt, G. D.; Rochester, C. H. (1983). "Adsorption of small molecules." In Adsorption From
Solution at the Solid-Liquid Interface. G. D. Parfitt ,C. H. Rochester, Eds.. New York,
Academic Press: 3-47.
Parham, R. A.; Gray, R. L. (1984). "Formation and Structure of Wood." In The Chemistry of
Solid Wood. R. Rowell, Ed. Washington, DC, The American Chemical Society. 207:
3-56.
Pierson, H. 0. (1993). Handbook of Carbon, Graphite,Diamond and Fullerenes. Park Ridge,
NJ, Noyes Publications.
Rael, J.; Shelton, S.; Dayaye, R. (1995). "Permeable barriers to remove benzene: Candidate
media evaluation." Journal of Environmental Engineering 121: 411-415.
Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M. (1993). Environmental Organic
Chemistry. New York, NY, John Wiley & Sons, Inc.
Schwarzenbach, R. P.; Westall, J. (1981). "Transport of nonpolar organic compounds from
surface water to groundwater: Laboratory sorption studies." Environmental Science
and Technology 15: 1360-1367.
Severtson, S. J.; Banerjee, S. (1996). "Sorption of chlorophenols to wood pulp."
Environmental Science and Technology 30: 1961-1969.
Stamm, A. J.; Millet, M. A. (1941). "The internal surface area of cellulosic materials."
Journal of Physical Chemistry 45: 43-54.
Steelink, C. (1985). Humic Substances in Soil, Sediment and Water. New York, John Wiley
and Sons.
Walters, R. W.; Luthy, R. G. (1984). "Equilibrium adsorption of polycyclic aromatic
hydrocarbons from water onto activated carbon." Environmental Science and
Technology 18: 395-403.
Wegner, T. H.; Baker, A. J.; Bendtsen, B. A.; Brenden, J. J.; Eslyn, W. E.; Harris, J. F.;
Howard, J. L.; Miller, R. B.; Petersen, R. C.; Rowe, J. W.; Rowell, R. M.; Simpson,
W. T.; Zinkel, D. F. (1992). "Wood." In Kirk-Other Encyclopedia of Chemical
Technology. J. I. Kroschwitz ,M. Howe-Grant, Eds.. New York, John Wiley and Sons.
162
I
-- _ -------- ---- I
~
--------------------~
Xing, B.; McGill, W. B.; Dudas, M. J. (1994). "Cross-correlation of polarity curves to predict
partition coefficients of nonionic organic contaminants." Environmental Science and
Technology 28: 1929-1933.
163
Chapter 6.
SORPTION OF NONPOLAR ORGANIC COMPOUNDS TO WOOD
164
~III-_
-X~I__~.-r_
I~-I-Y -.~Y-III~1~I-II_~L
LI-I-YI1--L^~-I~--t~ . -~-1.11 ._^ll~.^---^Pi --^~--^_
1111-._____
----^---
.-C--- _~_r~_l
Abstract
The sorption of nonpolar organic compounds to wood is important for determining the
groundwater transport of contaminants in aquifers of anthropogenic fill containing wood
wastes. Isotherms for benzene, toluene and o-xylene (BTX) sorption to Douglas fir and
Ponderosa pine chips were measured. Equilibrium wood partition coefficients for these
sorbates were 3 to 17 times less than values estimated on an organic carbon basis (i.e.,
Kwood = focKoc). A better model to predict wood partitioning accounted for the sorptivity of
the wood lignin and the lack of sorption by the cellulose component: Kwood =
flignmKgnm.
A
lignin-octanol free energy relationship with an equation of log K,,gnm = (0.72 ± 0.08) log Kow +
(0.08 ± 0.19) was developed using these lignin-normalized experimental
Kwood
values and
literature values of Khignm for other organic sorbates. The characteristic times of sorption
kinetics of all sorbate/sorbent pairs were on the order of 100 min for cm-sized wood particles.
Estimated characteristic diffusion times were in agreement with experimental values for both
physically hindered wood diffusion and homogeneous retarded diffusion models. The cleanup times of wood-containing aquifer solids is not limited by kinetic desorption from cm-sized
wood particles.
165
Introduction
Wood is an absorbent material which may be present as fill at many industrial sites.
During recent investigations at a former manufactured gas plant, wood was found in the
aquifer solids in the form of building rubble and chip waste from the gas manufacture process
(Chapter 5). The practice of burying fill was common at other manufactured gas plants
(Luthy et al., 1994), and likely was conducted at a wide variety of other industrial sites as
renovations were made or wastes were produced. Wood is also a significant component of
solid waste, accounting for up to 25% by weight of material at landfills which accept
demolition wastes (Niessen, 1977). Thus, groundwater transport of organic contaminants at
these sites may be influenced by the presence of wood in the subsurface solids. However, the
sorption of nonpolar organic compounds to wood has not been extensively studied and no
model exists to predict the influence of wood on contaminant transport.
The purpose of this study was to investigate the sorption of nonpolar organic
compounds to wood. The objectives of this experimental work were two-fold: (1) to
determine the equilibrium partition coefficients of structurally intact wood, and (2) to
determine the effects of particle size and sorbate hydrophobicity on the time scale of wood
sorption kinetics. The specific hypotheses that were tested while meeting these objectives are
developed in the following discussion of wood physiology and prior wood sorption studies.
Once a conceptual picture of wood sorption has been developed, the objectives will again be
summarized, noting our specific analytical approach.
166
-rar~-~-~.rau---- -t----~--i
- ---iurr-u~~,~lp-~-c------~--~--~I- ~'--UI-------
1
----
1111
Wood Physiology
Chemical Composition
Wood is a naturally occurring complex polymeric composite. The primary chemical
components of wood are cellulose and lignin. Cellulose is a carbohydrate formed from
repeating glucose units (Figure 6.1). The degree of polymerization of cellulose chains in
wood ranges from 1000 to 5000 units (Stamm, 1964). The hydroxyl groups make cellulose a
highly polar substance which is capable of hydrogen bonding. At the molecular scale, parallel
cellulose chains in wood cell walls form intermolecular hydrogen bonds to yield a rigid
crystalline structure. Cellulose constitutes a significant portion of wood, ranging from 3 to 47
weight percent in hardwoods and 40-44% in softwoods (Thompson, 1996).
Lignin is an amorphous three dimensional polymer which permeates the wood cell
wall and interstitial spaces (Pearl, 1967). It is formed by the enzymatic dehydrogenative
polymerization of phenylpropanoid monomers (Higuchi, 1980) (Figure 6.1). Hence, lignin-has
no specified chemical composition and is defined by its structural placement and function in
wood cells (Pearl, 1967). Unlike cellulose, lignin is hydrophobic in nature. The chemical
properties of lignin in wood are not known because isolation methods alter its structure (Pearl,
1967); however, isolated lignin exhibits a Hildebrand solubility parameter of 10 to 12
(cal/cm 3 ) 1 2 (Kopinke et al., 1995; Barton, 1983; Higuchi, 1980; Dellicoli, 1977) (c.f cellulose
solubility parameter of 14.5-16.5 (cal/cm 3 ) 1 2 (Barton, 1975)). By weight, hardwoods contain
16-24% lignin, and softwoods are 25-31% lignin (Thompson, 1996).
The remainder of the wood structure (about 20%) is composed of hemicellulose.
These are polysaccarides of 5 and 6-carbon sugars with short side chains (Parham and Gray,
1984). The irregular nature of these subunits limit the degree of hydrogen bonding that can
occur between chains, and hence limits the formation of a crystalline structure.
167
H
H
H
CH
2OH _I-O
-0
H
H
0
0
HO
H
H
0
HO
OH
H
H
H
OH
H
Cellulose
CH2 OH
CH 2 0H
CH2 OH
CH
II
CH
CH
CH
CH
CH
OCH 3
CH 0
OCH 3
OH
OH
OH
p-coumaryl
coniferyl
sinapyl
alcohol
alcohol
alcohol
Lignin Precursors
H
Predominant
C -- C-C
-0
Softwood
Lignin Linkage
OCH3
OCH 33
Figure 6.1. The molecular structure of cellulose and lignin polymers.
168
s~----~
-^------
II-~ .~I---~-~
1_--1
1_---_1^'~---~.l-~'~-PI.
-~-------*1~-1_-.1^1~_I -1_ ~111s~---~-_1
--1_~ __~I~
Physical Structure
Lignin and cellulose have a distinct arrangement in the wood cell walls. The basic
microstructure is a lignin-hemicellulose matrix with imbedded crystalline fibrils of cellulose
(Siau, 1984) (Figure 6.2). The cell walls also have a macrostructure composed of a primary
wall and 3 secondary walls (Figure 6.2). The interstitial region between the cells is referred
to as the middle lamella. The relative abundance of lignin and cellulose varies between the
cell wall layers, increasing in lignin content through the secondary wall to the middle lamella
(Figure 6.2).
The bulk wood structure is composed of regular layers of wood cells, referred to as
fibres. Fibres are the rigid cell wall structure remaining after autolysis of the protoplasm.
These fibres are oriented in the longitudinal direction of tree growth and enclose a hollow
lumen. Softwoods have an average longitudinal dimension of 3500 gtm (Siau, 1984), while
hardwood fibres have a shorter longitudinal dimension of 1000 to 1500 gtm (Stamm, 1964).
The ratio of longitudinal to radial dimensions of fibres is about 100 (Siau, 1984; Stamm,
1964) and the cell wall thickness is about 10 ptm. The lumina of adjoining cells are
connected by pits lining the sides and ends of the cell walls. These pits have a fibrous
covering to allow diffusive transfer of nutrients and water in the living wood tissue. Intercell
transfer is increased by the presence of ray cells oriented in the radial direction and connected
to the longitudinal fibres by pits. Ray cells only constitute 6 to 8% of the total wood porosity
(Stamm, 1964).
These features are present in both softwoods and hardwoods; however,
hardwoods contain additional structures known as vessel elements. Vessel elements are
continuous hollow channels, like straws, which transport sap.
Transport processes in hardwoods have been studied less extensively than softwoods
because of their more complex physical structure. The subsequent discussion, and our
experimental work, focus on softwoods because of their regular fibre structure. Softwoods
likely compose the bulk of the buried wood materials at industrial sites. Lumber in the form
of construction rubble, or lumber production wastes (sawdust, wood chips) used as sorbents,
may have been buried as fill and the primary class of building lumber is pine/fir/spruce. All
discussions are with respect to water-saturated wood.
169
Weight
Fraction
Lignin
Fraction
Total Wood
Lignin
3
trace
trace
S2
0.15
0.32
S,
0.52
0.38
P+
0.80
0.29
S
ML
Figure 6.2. The macrostructure and microstructure of the wood cell wall (Siau, 1984).
The primary (P) and secondary (S) cell wall layers and middle lamella (ML) are noted.
The lignin composition is for Scotch pine (total lignin fraction of 0.28) (Genco, 1996).
170
II.
-~ I- --------~-^-~-----~~I----
1--------- ---- ~
----..-- ~--I---
I
~~~~-~~~I-Y-^PIS~I(IC- PC
Sorption of Nonpolar Organic Compounds to Wood and Wood Components
Organic compound sorption coefficients have been measured for isolated wood
components. The magnitude of sorption to isolated cellulose and lignin reflects the respective
solubility parameters of these wood components. On an organic carbon basis, cellulose has
little measurable uptake of mono-aromatic compounds (Xing et al., 1994; Rutherford et al.,
1992; Garbarini and Lion, 1986) or chlorinated solvents (Rutherford et al., 1992; Garbarini
and Lion, 1986). By contrast, lignin exhibits linear isotherms for toluene, tricholoethylene
and chlorinated phenols (Severtson and Banerjee, 1996; Garbarini and Lion, 1986). Lignin
sorption coefficients are of a similar magnitude to those of soil organic matter when
normalized to the carbon content. The sorption of organic compounds to hemicellulose has
not been studied. These polysaccarides are similar in molecular composition to cellulose so,
like cellulose, hemicellulose would likely have little affinity for sorbing organic compounds.
From the above observations for cellulose and lignin, the sorption of organic
compounds to intact wood was hypothesized to occur by partitioning into lignin (Severtson
and Banerjee, 1996). Thus, a wood-water partition coefficient (Kwood) for a compound could
be predicted from the lignin content of the wood (fhignm) and the lignin-water partition
coefficient (Khgnm) of the compound, assuming all the lignin is accessible for sorption:
Kwood
(1)
=fhgmnKzhgnm
This hypothesis was tested for discrete wood fibres with varying lignin contents. A positive
correlation was observed between the fibre partition coefficients of chlorophenols and the
lignin fraction of the wood fibres (Severtson and Banerjee, 1996). The slopes of Kwood
v.
f
plots exactly matched independently measured Klignm values.
The sorption of benzene to intact wood particles in the form of sawdust was also
investigated (Rael et al., 1995). If all of the cell wall lignin was accessible in these wood
particles, the benzene partition coefficient should be predicted from Eq. 1. The estimated
partition coefficient was 16 mL/g (pine f~ignin = 0.26 (Petterson, 1984), benzene Khgnn = 60
(Xing et al., 1994)); however, the observed isotherm was highly nonlinear with a small
171
gnm
I~
(10-4 mL/g) Freundlich parameter (Rael et al., 1995). Eq. 1 greatly overestimated the benzene
wood partition coefficient suggesting that the assumption of lignin accessibility was incorrect.
The inability of Eq. 1 to predict
Kwood
for intact wood particles may be due to
differences in their structure relative to wood fibres. Most of the wood lignin is found in and
near the middle lamella region (Figure 6.2). If sorbates must first penetrate the cellulose-rich
secondary cell wall layer they may not be able to access all of the wood lignin. In contrast,
the lignin in the wood fibres of Severtson and Banerjee was exposed to the solution phase.
The discrete wood fibres were generated from wood chips by pulping, a process in which the
middle lamella is dissolved. Pulping also reduces the degree of cellulose polymerization
(Stamm, 1964), perhaps allowing sorbate access to lignin within the inner cell wall. The cell
wall fragments of the pine sawdust likely also had exposed secondary wall and middle lamella
lignin; however, deeper sorbate penetration into the whole lignin matrix may be inhibited by
outer regions of crystalline cellulose. Cell wall fractures may have been created as the
sawdust particles were generated by mechanical degradation, but no chemical degradation of
the wood matrix would have resulted.
Other data suggests that lignin-accessibility should not be hindered in intact wood
particles. First, visualization studies of wood impregnated with copper sulphate solutions
showed copper to be distributed throughout the cell wall matrix (Fengel, 1971). Second, the
model for diffusive transport through wood (described below) suggests that compounds
penetrate through the cell wall to some extent. Finally, the rigid crystalline structure of
cellulose fibrils may not be maintained under saturated conditions. At the fibre saturation
point, the stress at which wood failure occurs is at a minimum relative to wood with lower
moisture contents (Bodig, 1982).
Diffusion in Wood
The movement of compounds into the wood sorbent during sorption occurs by
diffusion. Thus, the kinetics of sorption reflect the characteristic mass transfer times of the
diffusional process. Two possible treatments of this diffusion process are considered. The
first model of the wood diffusion coefficient considers the diffusion hinderance of the
physical structure of the wood to compounds diffusing through wood particles. The second
172
-------------
--- -----Y----------
-------- - ----~"~~~P~~11111111--1
- -
,
------ "-~-----I
I
model overlooks wood cell structure and treats wood particles as homogeneous porous
sorbents.
173
Physically HinderedDiffusion
A model for compound penetration through saturated wood was developed by Stamm,
based upon the porous structure of wood (Stamm, 1964). A schematic picture of softwood
macrostructure is shown in Figure 6.3 for a bundle of fibres. There are two pathways for
diffusion through each wood fibre in the longitudinal or tangential (or radial) direction: (1)
diffusion through the continuous cell wall, or (2) diffusion through the lumen cavity.
Compounds move between the lumina of adjoining cells by diffusion through the overlapping
cell wall or through the interconnecting pits. As shown in the enlarged detail (Figure 6.3),
the pit structure is not an open aperture between adjacent lumina. A membrane composed of
cell wall tissue fills most of the pit chamber and is held in place by a porous fibre structure.
Pit diffusion may occur through the membrane itself, or through the membrane pores. Stamm
postulated that each of these pathways hindered the diffusion of solutes through a piece of
wood and referred to these hinderances as "resistances" to diffusion in the longitudinal,
tangential or radial direction.
A pictorial representation of Stamm's resistance model for diffusion through wood is
shown in Figure 6.4. By analogy to an electric circuit, the wood diffusion coefficient is
reduced relative to the solution diffusion coefficient by a factor equal to the reduced
conductance of the wood "circuit" relative to a circuit with no resistance (the solution).
Stamm's theoretical model was verified by measuring the conductance of wood blocks
saturated with a salt solution. The conductance of Western white pine was reduced relative to
the solution by a factor that was the same as the factor by which the experimentally measured
wood diffusion coefficient was reduced relative to the aqueous diffusion coefficient (Stamm,
1964; Burr and Stamm, 1947). Theoretical equations for the individual resistances depicted in
Figure 6.4 were developed with parameters derived from the physical structure of the wood
species. Details of the resistance calculations can be found in Stamm (1964) and Behr et al.
(1953). (They are also reproduced in Appendix C in a sample calculation specific to the
experimental Results of this chapter.) The resistance calculations do not account for
properties of the penetrant compound. Theoretically calculated ratios of wood-to-aqueous
diffusion coefficients show good agreement with experimental measurements for salts (Behr et
al., 1953).
174
IIIY~_^il-C.-.l-- ---_ll IIII~L-_
1.11~~~--1-1
-- IIII(CI^III~ liU19~~-~~~1__
(111_I..- -~..
-sPIC--^~_Y-*
Lumen
Cell
Wall
I~
~1111111~ 1111- ---111~-~11~1~1 1_- ~-~-~~~~--
Pit
Membrane
Pore
I1111
Pit
Enlarged Pit Structure
longitudinal
-
radial or tangential
Figure 6.3. Schematic representation of softwood physical structure with enlarged detail of
the interconnecting pit structure. The directions of fibre orientation are noted.
175
S Fibre Cavity
Ra
in Fibre Direction
• n - Number of Cells
per Unit Length
Rf
Pit
Cell Wall
Overlap
Rd
Membrane
Pore
Continuous
S Cell Wall
R TOT
Dwood
-
R
1
Daqueous
TOT
Figure 6.4. Pictorial representation of Stamm's resistance model for diffusion through
softwoods. The theoretical equation for the net reduction in the wood diffusion coefficient
relative to the aqueous diffusivity is also given (Burr and Stamm, 1947).
176
--- --------c~---
------ --------------
---~-"c~"'-~~~ ^^"-1~
r-- I~"P"~I----
I
The resistance model for diffusion in wood has also been applied to organic
compounds. Again, theoretical calculations of wood-to-aqueous diffusion coefficient ratios
predicted experimental values for lactose, glycerol, and urea (Stamm, 1964; Cady and
Williams, 1935). These organic compounds are highly polar and likely have little tendency to
sorb to wood tissue (e.g., urea octanol-water partition coefficient of 10.21 (Hansch et al.,
1995)). It is reasonable to assume that sorbing compounds would follow the same diffusion
pathways through the wood macrostructure as nonsorbing compounds. In the case of sorbing
compounds, diffusion through the cell wall tissue is likely retarded by equilibrium microscale
partitioning between the wood tissue and lignin. The cell wall tissue resistances, Ro, Re and
Rf are all inverse functions of the physical dimension (thickness or length) of the tissue
structure (Behr et al., 1953) (Appendix C). Microscale partitioning effectively lengthens the
diffusive path, and hence the resistances, R C,Re, and Rf for a sorbing compound would be
increased by a retardation factor that was a function of the lignin content of these tissues and
the lignin-water partition coefficient of the sorbate.
The effect of penetrant sorption on the overall ratio of wood-to-aqueous diffusion
coefficients is greater for tangential (or radial) diffusion than for longitudinal diffusion.
Conductance through the continuous cell wall is less than 4% of the overall wood
conductance in either the longitudinal or tangential direction (Burr and Stamm, 1947). The
resistance of the fibre cavities in the longitudinal direction is great compared to the pit
structures (Burr and Stamm, 1947), thus longitudinal diffusion coefficient would be relatively
unaffected by retardation of a sorbing penetrant molecule. In the tangential direction, the
resistance of the fibre cavities is much less important than passage between adjacent fibres for
which the pit resistance is about twice the resistance of the overlapped cell wall (Burr and
Stamm, 1947). Tangential diffusion coefficients could thus be affected by sorbate retardation
in lignin-rich cell walls.
Homogeneous Retarded Diffusion
The alternate model for diffusion through wood ignores the resistances of the cell
structure. Instead, wood particles are assumed to be homogeneous porous sorbents. Diffusion
occurs through the uncharacterized pore space in the wood. In the case of sorbing
compounds, the rate of diffusion is slowed by equilibrium microscale partitioning to the solid
177
sorbent. The effective diffusion coefficient is the aqueous diffusivity reduced by a retardation
factor which characterizes the partition process:
R =
+rKwood
(2)
where rw (g/mL) is the internal solid-to-water ratio, and Kwood (mL/g) is the partition
coefficient.
Scope of Investigation
The purpose of the equilibrium sorption study was to test the hypothesis that sorption
to intact wood is predicted by the fraction of lignin and the sorbate's lignin partition
coefficient (Eq. 1). Three sorbates (benzene, toluene, and o-xylene) were chosen for which
lignin-water partition coefficients were known (Xing et al., 1994). The sorbents were
Douglas fir and Ponderosa pine. The hypothesis that lignin partitioning is limited by
penetration through cellulose-rich cell wall regions was also tested with toluene sorption to
wood particles of differing sizes.
The second experimental objective was to determine the effects of particle size and
sorbate hydrophobicity on wood sorption kinetics. Wood particles were cut to include greater
fractions of the wood macrostructure, from cell wall fragments to hundreds of fibres. The
effect of sorbate hydrophobicity on uptake times was investigated using the three sorbates and
a single wood particle size.
Methods
Chemicals
Neat benzene, toluene and o-xylene were used as received from Alltech (Deerfield,
IL). Purified water (subsequently referred to as 18 Mn water) from an Aries water
purification system (Vaponics, Rockand, MA) was used to make aqueous solutions.
Acetonitrile was 'Baker Analyzed' HPLC solvent (J.T. Baker, Phillipsburg, NJ). Sodium
azide was from Fluka (Switzerland). Kiln-dried Douglas fir and Ponderosa pine wood was
obtained from Cambridge Lumber and Supply (Cambridge, MA). The wood was sanded
before use to remove surficial grit.
178
-r-------~-----i.-~*~-ri..-.~- ---~.I^-~^-I------r~--l-
..III -----------------
-,
..1..1~- I------------- -----I -I--------- -- ,--~C-
Equilibrium Sorption Isotherms
Wood partition coefficients were calculated from 5-point sorption isotherms. Douglas
fir sticks (2 cm x 0.1 cm x 0.1 cm, longitudinal x tangential x radial dimensions) and
Ponderosa pine chips (1 cm x 0.1 cm x 0.7 cm) were soaked in 18 Mf water containing
1 mM sodium azide biocide until they reached a constant wet mass. The minimum time of
contact was 18 days. The saturated wood particles were blotted with paper towel to remove
surficial water and transferred to 50 or 100 mL glass equilibration flasks. The flasks were
filled with an aqueous sorbate solution and sealed with glass stoppers to contain no
headspace. The stoppers were wrapped with teflon tape and clamped to prevent leakage. The
flasks were then wrapped with foil to minimize photodegradation. A wood-free control was
assembled in the same manner as the isotherm points to quantify loss mechanisms other than
sorption to wood. The flasks were inverted by hand several times daily to mix the solution
phase. Equilibration times were from 5 to 21 days (Table 6.1). At this time, the aqueous
concentrations were measured and the sorbed concentration calculated by difference from the
initial concentration, as detailed in Equations.
Sorption isotherms were determined for benzene, toluene, and o-xylene. Aqueous
sorbate solutions of various concentrations were made by diluting saturated stock solutions
with 18 Mn water in a glass syringe. Saturated stock solutions were made by equilibrating a
neat solvent phase of benzene, toluene or o-xylene with 18 MQ water in a separatory funnel.
The presence of the excess solvent phase ensured no compound loss from the saturated
aqueous solution between uses. Enough concentrated sodium azide solution was added to the
dilute aqueous solution in the syringe to give a final biocide concentration of 1 mM. The
syringe, containing glass beads, was shaken with no headspace to mix the solution. The
initial concentrations of the dilute aqueous phases for each isotherm point were measured
before transfer of the solution from the syringe to the equilibration flasks.
The dry weight and the water content of the wood particles was quantified at the
completion of the experiment. The wet wood particles were blotted with paper towel to
remove surficial water and weighed. The wood was then dried for 24 h at 103-105 0 C and
reweighed. The water content was the difference between the wet and dry weights of the
wood particles.
179
A minimum of 1 wood-containing flask and 1 wood-free control was also set up by
the above method for kinetic monitoring. These flasks were subsampled over time at time
intervals of 1 x 10n, 3 x 10n, and 8 x 10" minutes. The first value of n was 1 and sampling
continued until aqueous concentrations in the wood-containing flasks remained at constant
levels for several successive sampling points. The isotherms were measured at this time.
Aqueous concentrations were quantified by high pressure liquid chromatography
(HPLC). Thirty microliter aliquots were injected into a Hewlett Packard 1050 HPLC
equipped with a diode array detector. Compounds were eluted isocratically (85% acetonitrile/
15% 18 MQC water) through a 250 mm Adsorbosphere C,1 (5 jpm packing, Alltech) column at
a flow rate of 1 mL/min. Peak areas were quantified at a detection wavelength of 260 nm,
and background corrected for the solvent with a reference wavelength of 550 nm. External
standards were made up daily from the saturated stock solutions to calibrate the instrument
response.
Wood Sorption Kinetics
Toluene sorption kinetics were monitored for wood particles of various shapes with
toluene as the sorbate. Douglas fir and Ponderosa pine shavings were generated by rasping
wood. The sawdust was sieved and the fraction that was retained between 1000 to 1400 gm
screens was used for the time course experiment. This fraction also included some fines that
were stuck to the dry-screened particles. These particles were all broken cell wall fragments
with solution-exposed lignin. Larger particles, in the shapes of cubes, sticks and chips, were
also cut from the wood and rinsed of fines after being saturated with water. These larger
particles all had tangential and radial dimensions of 0.1 to 0.2 cm, except the chips which had
a radial dimension of 1 (pine) or 2 cm (fir). The cubes were about 0.3 cm in the longitudinal
dimension, less than the length of a softwood fibre. Therefore, longitudinal resistance to mass
transfer would not be important for these particles. The Douglas fir sticks and chips had
lengths of 2 cm and the Ponderosa pine sticks and chips had lengths of 1 cm. Both
longitudinal and tangential resistance to mass transfer would occur in these particles since
they were several fibre lengths long and 10s to 100s of fibre diameters thick.
The water-saturated wood particles were transferred to individual flasks as described
for the sorption isotherms. In this case only one aqueous toluene concentration was used for
180
Q___1~_
-.I-._
I_~II
LL-_I (-.ll^rr. -XIIIP -~L~ --..
lll~l~-L C-.IY
X-.~-. ~IIII I __.~_-~I
all of the wood particle shapes and the control flask. The stoppers were fitted with ground
glass stopcock joints to permit subsampling with a minimum of headspace turnover. Aqueous
concentrations were analyzed by HPLC at logarithmically spaced intervals as described for the
kinetic isotherm monitoring.
All data from kinetic monitoring are reported on plots of peak area versus time. The
HPLC response factor for the calibration standards did not vary over the duration of the time
courses. Therefore, peak area was an accurate surrogate for aqueous sorbate concentration.
The term aqueous concentration is used in place of peak area in the discussion section.
Equations
Sorption Isotherms
Sorption isotherms were developed from mass balances on the equilibration flasks. At
equilibrium, the total mass of sorbate in the initial solution, Mto,, was equal to the sorbed
mass, Mwood, the mass remaining in solution, MwVr, the mass in the headspace, Malr, and the
mass dissolved in the water-filled wood porosity, Mdiss'd,:
water +Mair
tot =Mwood +
(3)
+ Mdd
The sorbed compound concentration was calculated by difference from the initial
concentration as follows:
C
S
,
M
-K.
-c
I
W
MC
Vw
(4)
f
where C, (pg/g) is the sorbed concentration; V (mL) is the bulk water volume; M (g) is the
dry mass of wood; C, (gg/mL) is the syringe transfer-corrected initial aqueous concentration;
C, (gg/mL) is the final aqueous concentration; KH (mL/mL) is the dimensionless Henry's Law
constant; Va (mL) is the headspace volume, and M, (mL/g) is the mass normalized volume of
water-filled wood porosity.
All of the variables in Eq. 4 were measured except Henry's Law constants. KH values
were taken from Schwarzenbach et al. (1993). The syringe transfer-corrected initial aqueous
concentration was the sorbate concentration measured in the dilution syringe and multiplied
181
by a factor of 0.92. The sorbate concentrations for 5 wood-free controls were measured
immediately before and after transfer to equilibration flasks. The aqueous concentrations in
the flasks were found to be only 92 ± 2% of the syringe concentrations. On average, 8% of
the sorbate mass was lost for these wood-free controls during syringe transfer, therefore, a
constant 8% sorbate loss was also assumed for each of the isotherm flasks. The volume of
headspace was determined by the difference in the mass of the equilibration flasks after
assembly and prior to isotherm analysis. Losses of sorbate to the 0.2 to 0.6 mL headspace in
the flasks were insignificant. Finally, the mass normalized volume of water-filled porosity
was calculated by dividing the volume of water lost upon drying the wood particles by the
wood dry weight.
Wood partition coefficients were calculated from plots of sorbed concentration versus
aqueous concentration. A linear regression was used to fit a linear sorption isotherm to the
data:
C
Kwood
(5)
=
w
where Kwood is the wood-water partition coefficient. The linearity of the fits was also verified
by fitting the log-transformed data to the log-transformed Freundlich equation:
log C = nlog C +logK
(6)
where n is the Freundlich exponent and Kf is the Freundlich coefficient. If the value of n was
not significantly different than 1, a linear sorption isotherm was assumed to apply.
Sorption Kinetics
The characteristic time scale for a mass transfer process is the time at which half of
the equilibrium sorbed mass has been removed from solution. The mass transfer times that
were experimentally measured in the kinetic time courses are referred to as t112 values in order
to distinguish them from characteristic times that are theoretically calculated. Experimental
t2,, values were calculated by first fitting a linear equation to the changing aqueous
concentrations (before the equilibrium plateau) on a linear-log plot of peak area v. time. The
182
;--~--~II-~--~ 1U---^---- ---~---^---CII~'~^'~ I~VI~~-~'~'-'~ll---l
~.._-~
-XI~_I~-~-~-YLI-.
I11~-I-~---_I-~-I_--L-- C~
I^L1- ..1----~
_
_ Il--IFII_
time at which the peak area of the best-fit line was equal to the midpoint of the first measured
peak area and the equilibrium plateau was the t,,, value.
Characteristic time scales for mass transfer processes were also theoretically
calculated. The important sorptive uptake mechanism for the wood was diffusive mass
transfer. The characteristic time at which half of the equilibrium sorbed mass has been
removed from solution was assumed a function of a characteristic diffusion length scale and a
diffusion coefficient:
t
12
1
Def
(7)
where 1 (cm) is the characteristic length and Deff (cm2/s) is the mass transfer process-specific
diffusion coefficient.
Results and Discussion
Equilibrium Sorption Isotherms
Wood-water partition coefficients
Benzene, toluene and o-xylene (BTX) were sorbed by particles of Douglas fir and
Ponderosa pine wood. The aqueous concentrations of these compounds decreased to a
significantly greater extent in flasks with wood particles than in controls containing no wood
(Figure 6.5). Compound mass recovered from successive desorptions of the wood particles
was greater than 93% of the sorbed mass calculated by difference between the initial and final
aqueous concentrations. Thus, solute concentrations in the wood-containing flasks did not
decrease by degradation by bacteria or fungi introduced into the solution with the wood. The
decrease in aqueous concentrations, relative to the controls, was greater than the decrease
calculated for dilution into the water-filled porosity alone. For both woods and all sorbates,
the fraction of sorbate mass remaining in the aqueous phase at equilibrium (f,) was less than
would have remained had the bulk sorbate concentration been diluted by diffusion into the
water-filled pore spaces of the wood particles. For example, the volume of water in the wood
particles for the Ponderosa pine-benzene couple was 13% of the water volume in the
183
7~A
550
500
*
Control
o
A
Chips
Chips
0
450
0
400
350
300
250
B
20
O
a
,,
0
100
200
300
500'
400
2000
6000
4000
8000
10000
Time (min)
500
-
0
450
DA
400
A
A
350
300
250
200
0
0
A
Control
Chips
Chips
i i i ,.
EA
i
i
i,
i
i
,
I
.
i
i
.
t
ii
1000
10000
Time (min)
Figure 6.5. Change in aqueous toluene peak area as a function of time for duplicate flasks
containing Ponderosa pine chips.
184
.PI_ .~--~1...-11~.~---..-3^.-..~---..--.~.^-*11~-1--1__-1
I
-------~II~-"~~
,-~
------
~-~--^I~~~~~ .~~--CI~~I~I~
~^I~.~
I~llll~l-~-~C
equilibration flask. The concentration of benzene in the bulk solution would be diluted to
88% (1/1.13) of its initial concentration solely by diffusion into the benzene-free pore space
water. The observed fraction of benzene mass in the aqueous phase at equilibrium was 54%
of the initial mass, indicating that wood uptake in addition to pore space dilution had
occurred. An effect of sorbate hydrophobicity on f, was observed for Douglas fir for which
the same solid-to-water ratios were used for each of the isotherms (Table 6.1). The fraction
of mass remaining in the aqueous phase decreased with increasing compound hydrophobicity,
consistent with sorptive uptake of alkylated benzenes by wood.
The sorptive uptake of toluene was also not limited by sorbent particle size. The
fraction of toluene mass remaining in the aqueous phase at equilibrium was 0.56 for
Ponderosa pine cubes and sticks and 0.51 for cubes. The equilibrium "accessibility" of the
partitioning phase appeared to be independent of wood particle size.
Sorption isotherms were plotted for each of the sorbates and each of the softwoods
(Figures 6.6 and 6.7). Errors in the aqueous concentrations were determined from the
variability in peak areas of replicate injections and the error in the HPLC sorbate response
factor. Errors in aqueous concentration measurements were propagated through Eq. 4 to give
the error in solids concentrations. Final control concentrations were 90 ± 3% of the initial
transfer-corrected concentrations. No corrections for leakage losses from wood-containing
flasks were made.
Most of the sorption isotherms had Freundlich exponents which were not significantly
different than 1 when the log-transformed data was fit to Eq. 5. The exceptions were toluene
sorption to Ponderosa pine (0.9 ± 0.04) and o-xylene sorption to Douglas fir (0.8 ± 0.1). The
linear plot of the Douglas fir-o-xylene isotherm also had a nonzero intercept. Both of these
isotherm characteristics likely resulted from analytical limitations. The equilibrium aqueous
phase concentrations of o-xylene that were less than 2 pg/mL were near the HPLC detection
limit. For these 3 data points, the fraction of compound mass in the water was only 10%
while, f, was 0.39 ± 0.03 for the other data points. If the aqueous phase concentration was
underestimated, the solids concentration would be correspondingly overestimated when
calculated by difference from the starting concentration. This shift in concentrations would
lead to an apparent nonzero isotherm intercept and a nonlinear Freundlich isotherm fit.
185
Table 6.1. Experimental conditions and partition coefficients for equilibrium isotherms.
Benzene
Toluene
o-Xylene
DOUGLAS FIR
Equilibration time
8d
9, 22 dt
5, 15 dt
Average f, (n = 8)
0.63 ± 0.01
0.57 + 0.02
0.39 ± 0.03
0.04
0.04
0.04
17 ± 0.8
23 ± 1
34 ± 4
60
80
110
Porosity dilution:
Kwood (mL/g)
Klgnin
(mL/g)
PONDEROSA PINE
Equilibration time
12 d
9d
14 d
Average f, (n = 9)
0.54 ± 0.03
0.45 ± 0.05
0.48 ± 0.03
0.13
0.09
0.06
8.8 ± 0.3
18 ± 2
33 ± 1
30
60
110
Porosity dilution:
K d (mL/g)
KhIgnm (mL/g)
PHYSICAL CONSTANTS AND PARTITION COEFFICIENTS
Kow
(Miller et al., 1985)
Koc
(Schwarzenbach et
al., 1993)
Kwood = focKo
(mL/g)
135
450
1350
54
180
550
30
60
110
t Second value is duplicate measure of same isotherm flasks that were measured at the
earlier time point.
: Porosity dilution = (Total volume of water in wood particles)/(Bulk volume of water in
flask)
186
-
I
I
II
I
4000
3000
2000
1000
0
Aqueous Concentration (jtg/mL)
2500
2000
1500
1000
500
0
20
40
60
80
100
120
Aqueous Concentration (Rig/mL)
1500
1000
500
0
5
10
15
20
25
30
35
40
45
50
Aqueous Conc entration (Rjg/mL)
Figure 6.6. Ponderosa pine sorption isotherms;. Partition coefficient values are reported in
Table 6.1.
187
8000
6000
4000
2000
0
100
200
300
400
500
Aqueous Concentration (ig/mL)
4000
3000
2000
1000
0
20
40
60
80
100
120
140
Aqueous Concentration ( ig/mL)
1500
1000
500
0
10
20
30
40
50
Aqueous Concentration (pg/mL)
Figure 6.7. Douglas fir sorption isotherms. Partition coefficient values are reported in
Table 6.1.
188
.~~lrr^l-~...~----~-~I
- -l ~--~u~--~-rr------- C--'~"~~~'---sc-- ~_~,~_ ..~1~_ --- -- ----------------- I- - ------
Wood partition coefficients,
Kwood,
I
I
were calculated from the slopes of the sorption
isotherms (Table 6.1). The partition coefficients for Ponderosa pine approximately doubled
from benzene to toluene to o-xylene. This trend was consistent with the hydrophobicities of
these compounds: the octanol-water partition coefficients approximately triple with the
addition of each methyl group to benzene. The wood-water partition coefficients for Douglas
fir also increased with hydrophobicity, but not in regular multiples. Benzene and toluene had
about the same Kood, while the o-xylene Kood was about 50% greater than for toluene. Both
of the species of wood did exhibit partition coefficients of similar magnitude.
The lignin-normalized partition coefficients for the wood-BTX pairs were compared to
literature values for these sorbates. The range in lignin contents of softwoods is 25-31% by
weight (Thompson, 1996). This not a wide enough range to calculate Kignm for undigested
wood from a plot of Kwood
v.
flignm
(Eq. 1). Rather, we normalized our experimental Kwood
values directly by dividing by the fractional lignin content. Douglas fir and Ponderosa pine
are both about 30% lignin by weight (Parham and Gray, 1984). The resultant Kignm, partition
coefficients are reported in Table 6.1.
The calculated K,gnm values for benzene, toluene, and o-xylene showed good
agreement with literature partition coefficients for isolated lignins. Xing et al. (1994) reported
lignin-water partition coefficients for an organosolv lignin of 60, 140, and 330 mL/g for
benzene, toluene, and o-xylene, respectively. Kh~gnm values for these compounds were 30, 90,
and 180 mL/g for an alkali lignin. The toluene partition coefficient for an alkali pine lignin
was 100 mL/g (Garbarini and Lion, 1986). If the Kwood values for Douglas fir and Ponderosa
pine were estimated with alkali lignin partition coefficients, they would have agreed with our
measured values to within a factor of 2 or less.
Predictions of wood partition coefficients on an organic carbon basis greatly
overpredict our experimental values. Traditionally, partition coefficients for soils and
sediments are estimated from the sorbent fraction organic carbon, foo, and an organic carbonnormalized partition coefficient, K,, (i.e., Kd = focKo) (Karickhoff et al., 1979). This method
was applied to Douglas fir and Ponderosa pine wood using Koc values estimated from an
octanol water partition coefficient linear free energy relationship (Table 6.1) (Schwarzenbach
189
s.
et al., 1993). The average carbon content of wood is 50% (Wegner et al., 1992) so a value
of 0.5 was used for both woods. The calculated partition coefficients, Kd, overestimated the
measured values by 2 to 3 times for benzene, 5 times for toluene and 7 to 8 times for
o-xylene (Table 6.1). This result is not surprising given that only the lignin polymeric wood
component has been demonstrated to sorb organic compounds (Xing et al., 1994; Garbarini
and Lion, 1986). The carbon content of the wood used to estimate Kd included both lignin
carbon and cellulosic carbon which does not contribute to sorption. Douglas fir and
Ponderosa pine are about 30% lignin by weight (Parham and Gray, 1984). Since lignin is
about 50% carbon by weight (Xing et al., 1994; Garbarini and Lion, 1986), the true sorbing
foc should be 0.5x0.3 = 0.15, or a factor of 3 less than the foc used to compute the Table 6.1
Kd values. Accounting for the sorptive capacities of the wood polymers, the wood partition
coefficients were estimated to be 8, 30, and 80 for benzene, toluene, and o-xylene,
respectively. These refined partition coefficient estimates are much closer in value to the
wood partition coefficients measured for Douglas fir and Ponderosa pine, but the ratio of
calculated-to-measured values is not constant, suggesting that lignin has a different octanolwater free energy relationship than natural organic matter.
Linear Free Energy Relationshipfor Lignin-Water Partition Coefficients
A linear free energy relationship (LFER) relating lignin and octanol water partition
coefficients was developed. Such a free energy relationship is constructed assuming that the
same interaction occurs between the sorbate and lignin as between the sorbate and octanol.
Lignin exhibits noncompetitive compound uptake with a low (1.8 - 2.6 kcal/mol) heat of
sorption (Severtson and Banerjee, 1996), suggesting that organic compounds sorb to lignin by
a partitioning mechanism, such as compounds distribute between octanol and water. This
argument has been used to develop LFERs for natural organic matter sorption (Chiou et al.,
1979). Organic matter macromolecules in an aqueous solution are free to adopt
configurations that contain hydrophobic pockets into which organic solutes can partition. In
contrast, the function of lignin is to impart rigidity to the wood structure (Parham and Gray,
1984). This may limit the extent to which organic molecules can partition into this polymer.
As a note, Severtson and Banerjee (1996) measured heats of sorption and sorptive
190
I --- ----------
II
--
I
competitiveness on wood fibres. These fibres were generated in a pulping process which may
have chemically degraded the lignin so that it could more easily form hydrophobic pockets for
sorbate molecules. However, intact lignin partition coefficients measured in this experiment
compared well with values for highly digested alkali lignin (Xing et al., 1994). This
agreement suggests that the tertiary structure of the lignin in intact softwoods does not alter
its sorptivity. The linearity of the Douglas fir and Ponderosa pine isotherms was more
evidence suggestive of partitioning between lignin and water.
Lignin-water partition coefficients, Kgnm, were plotted versus octanol water partition
coefficients (Figure 6.8). The regression of the free energy relationship had a slope of
0.71 ± 0.08 and an intercept of 0.08 ± 0.19. If the lignin LFER is converted to a carbon
basis by dividing by the 50% carbon content, the resultant equation is:
logK,, = 0.71 logKow + 0.38
(8)
This organic carbon-normalized LFER is similar to the free energy relationship for substituted
benzenes developed by Schwarzenbach and Westall (1981). The equation for natural aquifer
sediment organic matter was log Koc = 0.72 log Kow + 0.49. The similarity between the free
energy relationship slopes of lignin and sediment organic matter is consistent with the
diagenic formation of organic matter. Lignin and other plant materials are precursors for
natural organic matter in terrestrial systems (Kumada, 1987).
There was remarkable agreement between the Ki,,, values for several forms and
sources of lignin. The individual LFER slopes for Ponderosa pine (0.6) and isolated lignin
(lignin O and A slopes = 0.7) were not significantly different than for the whole data set;
however, the Douglas fir data had a regression slope of 0.3. The data sets for individual
lignins were limited to only three compounds. Thus, it is not clear whether lignin partition
coefficients for Douglas fir were distinct from the other wood and lignins. A slope of 0.3
suggests highly unfavorable interactions of the sorbates with this wood, relative to octanol.
Certainly the saturated Douglas fir was more rigid than the Ponderosa pine, if the tertiary
structure of intact wood is important to sorption. However, as previously noted, Kignm
measurements for compounds with a wider range of Kow values is needed to conclude if there
191
1000
*
D. Fir, this thesis
O
A
A
O
0
P. Pine, this thesis
Xing et al., lignin O
Xing et al., lignin A
Garabini & Lion
Severtson & Banerjee
-
Stamm and Millett
O.-
100 'oE
10
."
= (0.71 + 0.08) log Kow + (0.08 + 0.19)
log K,
r = 0.83
PH
10
BZ
TCE TL
100
OX DP
1000
TP
10000
KOW
Figure 6.8. Lignin-octanol linear free energy relationship. Experimental and literature data
are plotted for phenol (PH), benzene (BZ), trichloroethylene (TCE), toluene (TL), o-xylene
(OX), 2,4-dichlorophenol (DP), and 2,4,5-trichlorophenol (TP). The solid line is the best fit
linear regression with 95% confidence intervals denoted with dotted lines.
192
-
__~_____,,
_ __,
I
-_ _II
are significant sorptive differences between various species of wood lignin. The low slope of
the individual Douglas fir LFER may have simply been skewed by the analytical uncertainty
in determining the o-xylene Khignm
The intercepts of the individual LFERs varied among the different lignin types. These
variations may reflect compositional differences between wood species or extraction methods.
By analogy, the aromatic carbon content of natural organic matter allows fine tuning of
organic carbon partition values (Chin et al., 1997). Since lignin does not have a well defined
molecular structure, the carbon content of the lignin may explain the data spread. For
example, lignin O had a carbon content of 65.8% and an intercept of 0.18 and lignin A had a
carbon content of 57.1% and an intercept of -0.02. If the carbon-normalized partition
coefficient was the same for each of these lignins, the lignin-water partition coefficients
would be related:
logK,,,= logK
logK,,,
-C
+ log fo2
(9)
+ 0.06
If lignin O were lignin' and lignin A were lignin' in Eq. 9, the intercept for lignin O would
be greater than the intercept for lignin A by 0.06 as a result of their differing carbon contents.
The observed intercept difference of 0.2 was greater, suggesting that more subtle molecular
differences may exist between the two lignin types. As a whole, Klignm values in Figure 6.8
showed remarkable consistency despite the physical structure (whole wood v. isolated lignin)
and the source (organosolv, alkali, Kraft) of the different lignins.
(Note that a prior lignin-octanol LFER published by Severtson and Banerjee (1996) is
erroneous. First, they did not report the source of Kow values. In a compilation of values
(Mackay et al., 1992), the reported log Kow for 2,4-dichlorophenol generally fell between 3.1
and 3.2 and the log Kow for 2,4,5-trichlorophenol was between 3.7 and 3.8, ranges both
greater than the values used by Severtson and Banerjee. Secondly, they plotted the carbonnormalized lignin-water partition coefficients reported by Garbarini and Lion (1986) without
noting the change in nomenclature. Finally, the Kignm values measured by Severtson and
193
_ ~
Banerjee in their experiment do not appear to be plotted correctly in their LFER. The plotted
log Klignm for 2,4-dichlorophenol, with a value greater than 3, is too great even if it also was
carbon-normalized. The Kgn,, value for 2,4-dichlorophenol was reported in the text as 300 by
two methods of determination. With an average carbon content of 50%, the carbonnormalized value would be 600, or 2.8 on a log scale. The 2,4,5-trichlorophenol datum was
similarly high. The data of Severtson and Banerjee and Garbarini and Lion were plotted with
corrected values in Figure 6.8. The individual LFER equation for this data was still similar to
Severtson and Banerjee's lignin free energy relationship: Figure 6.8 slope of 0.9 and intercept
of -0.32; Severtson and Banerjee (Severtson and Banerjee, 1996) slope of 0.95 and intercept
of -0.48. The LFER developed in Figure 6.8 is a better predictor of lignin-water partition
coefficients from octanol-water partition coefficients because it incorporates several lignins
from several sources and a wider range of sorbate hydrophobicities.)
The results of these sorption experiments with Douglas fir and Ponderosa pine suggest
that the sorption of nonpolar compounds to structurally intact softwoods can be predicted by
knowing the lignin content and a lignin-octanol LFER. By extension, it is reasonable to
assume that hardwood sorption coefficients can also be predicted from the lignin content.
Compositional differences are known to exist between hard- and softwood lignins due to
varying ratios of the phenylpropanoid precursors (Parham and Gray, 1984); however, the
softwood lignin LFER is probably still applicable to hardwoods. The chemical composition
of the three monomer compounds are similar, varying only in the number of methoxyl groups.
There are about 0.95 methoxy groups per phenylpropanoid unit in hardwoods and 1.3 to 1.7
methoxyl groups in hardwoods (Genco, 1996).
We note again that the prediction of K,,ood values from lignin-dominated partitioning is
applicable only to fully saturated wood. The rate of diffusion of a wetting front into wood is
significantly slower than for sorbent molecules (e.g., 18 days to saturate Ponderosa pine chips,
less than 9 days to equilibrate toluene). This may be a possible explanation for the low
sorptive uptake of benzene observed by Rael et al. (1995), although their equilibration time
was greater than a day, the time at which our pine shavings were thought to be saturated.
194
I~-^-~~ IIII..LIII__T
-- ~~~~~--^---I*_LL__ __XIIIIIII
Il-^IC-~^-~l~-------------P
~-i~u3~c~ur*~-~i~--~
--m~-r~L.,I-p--mop
I
Kinetics of Wood Sorption
Experimental t,,2 Values
Compound peak areas in the Douglas fir and Ponderosa pine-containing flasks
decreased with time, with apparent bimodal kinetics. A representative time course is shown
in Figure 6.5 for toluene (refer to Appendix D for the complete set of all sorbate-sorbent
pairs). A similar pattern was also observed for benzene and o-xylene. Sorbate peak areas in
wood-containing flasks decreased rapidly over the first 1000 min (17 h) of equilibration,
followed by a slower decrease in aqueous concentrations. The time courses were developed
by repeated monitoring of compound concentrations in single flasks, rather than sacrificing
new flasks at each time point. Over time, the headspace in the flasks increased as subsamples
were removed for analysis. Between sampling episodes, the volatile sorbates would partition
into this headspace and may have been lost while opening and closing the flasks at each
sampling point. In order to verify that sorption equilibrium was reached before 10 000 min,
the Douglas fir toluene and o-xylene sorption isotherms were measured twice, first after
13 000 and 7200 min, respectively, and again after 31 000 min and 21 000 min, respectively.
Since there was no change in the aqueous concentrations of these two sorbates between the
first and second sampling times, and sorbate concentrations in wood-free controls also began
to noticeably decline at time scales approaching 10 000 min, we concluded that the apparent
slow kinetics likely reflected loss mechanisms other than sorption. Therefore, the discussion
of wood sorption kinetics focusses on the initial, fast kinetics process when the decrease in
aqueous concentrations were dominated by sorptive uptake.
Time courses were plotted for wood particles of varied dimensions to determine the
rate limitations to sorptive uptake by this sorbent (Figures 6.9 and 6.10). Wood shavings
were hypothesized to have the fastest time to sorptive equilibrium because the lignin in these
broken cell wall fragments was accessible to the aqueous solution. Indeed, the fastest uptake
of toluene by Ponderosa pine and Douglas fir was observed for the shaved particles
(Figures 6.9 and 6.10, Table 6.2). After 30 min of equilibration, toluene peak areas in the
aqueous phase remained constant (Douglas fir, Figure 6.10) or decreased at the same rate as
the wood-free controls (Ponderosa pine, Figure 6.9), indicating that sorptive uptake was
195
i
J
J
0
O
A
+
*
500
450k 0
Shavings
Cubes
Sticks
Chips
Control
0
0
400
350
A 0-
300
250
S
00
-o0
.
.
O
02
.
.................
200
0
100
200
iiiii
400
300
. . .
.*..
, , .
* ,
*
, *,,,
.
06 2000 4000 6000
iI
8000 10000 12000
Time (min)
0
O 0
0
0
Shavings
0
A
Cubes
Sticks
+
Chips
*
Control
A
0
0
0
A4
O0
o
l
l
I
I
I
I,,
I
,
1000
,
I I,...,
,...
I
I
10000
Time (min)
Figure 6.9. Decrease in aqueous peak areas of toluene as a function of time for varied
sizes of Ponderosa pine wood particles.
196
-v-- -r~l~^
r -------- ~--r~-~----
I
~I- ......
~------------
-L----~----------
~.-
550
*
o
500
Control
Shavings
Cubes
o
450
A
Sticks
•
Chips
400
350
0
300
A
0
9
0
o
0o
250
200
1CCA
i*
i
200
100
i
I
300
Lf/~
5000
15000
10000
Time (min)
250
o
Shavings
0
Cubes
200
A
*
Sticks
Chips
15 0
.
1
.
.
.
10
'
'
100
1000
.
.
. .
10000
Time (min)
Figure 6.10. Decrease in aqueous peak areas of toluene as a function of time for varied
sizes of Douglas fir wood particles.
197
I
not important after this time. In contrast, sorptive uptake of toluene by the other wood
particles remained an important process for greater than 1000 min. The apparently greater
equilibrium peak areas for the Douglas fir and Ponderosa pine shavings, relative to the other
wood particles, was a consequence of differing solid-to-water ratios. The solid-to-water ratio
in the shavings flasks were lower by a factor of 2 than the other flasks, and hence the fraction
of compound mass in the aqueous phase at equilibrium was greater for the wood shavings
than the other wood particles.
The Ponderosa pine cubes, sticks and chips all had similar time scales at which 50%
of the sorbed mass had been removed from solution (Table 6.2). The values reported in
Table 6.2 were from single flasks. The variability among single replicate measures of t 12, was
quantified by monitoring duplicate and triplicate kinetic time courses during the sorption
isotherm experiments. The time scales to reach 50% uptake for toluene and benzene were
between 90 and 120 minutes and 30 and 90 minutes, respectively (Table 6.3). From the
range in t 1/2 values calculated from the multiple time course measurements, individual single
point estimates of t1, 2 were thought to vary by 20 to 50%. Thus, the toluene uptake times for
cubes, sticks, and chips were likely not significantly different from one another. Because
there was little difference in the uptake times for these particles, and these particles all had a
common tangential length scale of 1 to 2 mm, we concluded that the toluene uptake kinetics
in Ponderosa pine cubes, chips and sticks was limited by tangential diffusion. If longitudinal
diffusion had been important, the uptake times for the sticks and the chips would have been
25 times greater than for the cubes according to Eq. 7 since these particles had longitudinal
lengths 5 times greater than the cubes.
The time scale of the fast kinetic process was determined for Douglas fir with data
points collected in the first 1000 min of equilibration (Table 6.2). Only the cubes appeared to
have reached constant aqueous peak areas by this time (Figure 6.10). The aqueous toluene
concentrations for the sticks and chips continued to decrease beyond this time. The sticks
were known to reach equilibrium sorptive uptake within 13 000 min from the isotherm
experiments so the continued decrease in toluene concentrations in the kinetic time course
was probably a result of headspace losses. The time scales for the fast kinetic uptake by the
198
~~-I-
L
~
---
l-
--
--.
.II
Table 6.2. Kinetic uptake of toluene by wood particles of various shapes. Reported values
are the times, in minutes, at which half of the sorbed mass was taken up by the particles.
Douglas fir
Ponderosa pine
Shavings
Cubes
Sticks
Chips
2
150
100
270
< 8
60
50
120
199
Table 6.3. Kinetic uptake by Ponderosa pine chips.
t1 /2
Benzene
Toluene
o-Xylene
30
80
90
90
120
110
0.7
1.1
1.1
1.06 x 10 5
9.4 x 10-6
8.5 x 10 -6
7
12
18
(min)
rbulKd
D, (cm 2 /s)t
R
tCalculated from Hayduk and Laudie (Schwarzenbach et al., 1993).
Table 6.4. Kinetic uptake by Douglas fir sticks.
Benzene
Toluene
o-Xylene
t1/2 (min)
62
100
250
rbulkKd
0.6
0.8
1.2
9.4 x 10-6
8.5 x 10- 6
19
27
D, (cm 2/s)t
R
1.06
x
10'
14
tCalculated from Hayduk and Laudie (Schwarzenbach et al., 1993).
200
~--~
.-.. ------------,--- -.~~-- ----
-----------
-- I------------~...I
.. -c~--'---~-ll-------
I~~l-~~~-s~.
~xl"l~-^IIYs ~~~~^III------'~~~~~~"~'~I~CIIx~"l~
cubes and sticks were not significantly different within the variability of individual t1/2 values,
but the chips may have exhibited longer toluene uptake times than the smaller particles.
The effect of compound hydrophobicity on the tl, 2 values was investigated by
comparing the uptake times for each of the isotherm sorbates. In order to compare the
kinetics of sorbing compounds, the product of the bulk solid-to-water ratio, rbulk, and the
partition coefficient, Kd, must be constant (Wu and Gschwend, 1988; Crank, 1979). This
product was near 1 for each of the sorption isotherms and so intercomparisons were made
between the uptake times (Tables 6.3 and 6.4). There was no difference in the t1 /2 values for
benzene, toluene, and o-xylene and Ponderosa pine. There appeared to be only a slight effect
of sorbate hydrophobicity on t1 /2 times for Douglas fir because o-xylene exhibited slightly
slower uptake kinetics than benzene and toluene.
CharacteristicDiffusion Times - Physically Hindered Diffusion
The measured isotherm tl/2 values were compared to characteristic times estimated for
the physically hindered and the homogeneous retarded diffusion models. The characteristic
times were calculated with Eq. 7 using length scales and diffusion coefficients tuned to our
experimental system. First, the characteristic length scale was half the tangential thickness of
the wood particles, 0.05 cm. This length scale was used for the homogeneous porous sorbent
model because the ratio surface areas in the longitudinal and tangential directions are
proportional to the area ratios, and thus wood uptake would be dominated by diffusion in the
tangential direction. Secondly, retardation factors were calculated from Eq. 2 (Tables 6.3 and
6.4). The internal solid-to-water ratio was determined from the moisture content of the wood.
The values were 0.78 g/mL for sorption of all sorbates to Douglas fir, and 0.65, 0.63 and
0.52 g/mL, respectively, for benzene, toluene and o-xylene sorption to pine. The bulk wood
partition coefficients (Table 6.1) were used for Kwood. Detailed calculations of the diffusion
coefficients follow.
Wood diffusion coefficients were first calculated with Stamm's resistance model,
accounting for the physical structure of the wood. The individual resistance values for
tangential diffusion were calculated with values for Douglas fir from Behr et al. (1953)
(Appendix C). The resistance model is depicted in Figure 6.11 a with values reported as
201
0.036
0.015
98
300 X
300 X
0. 0
4.45
39
3.90.00022
61
8
18
152
15.2
98
7
2
5.
5.2
6.9
60.69
0.039
0.0152
(b) Stamm's diffusion model with
retarded cell wall diffusion
(a) Stamm's diffusion model
Figure 6.11. Calculation of tangential Douglas fir conductance with Stamm's resistance
model for wood. Conductance values from Behr et al. (1953) are noted in the ovals. The
conductance of each branch with multiple conductances is noted in bold at each node, and
the total conductance is noted in bold at the bottom node. Percentage conductances of each
branch are noted in italics.
202
^.^-*II. ~--~------- ---
L~-------~--l
-L~---^-I__IIIXII--..-ll
LI
-ili IYU(_
I-- --
^I --II~ _
I----..
-tlllll-~---l___l_____-~~ 1 111 1~~1_____1_ %
relative (to aqueous solution) conductances. The overall relative conductance of this "circuit"
is 0.04, that is, the tangential wood diffusion coefficient is 0.04 times the aqueous diffusion
coefficient. This tangential diffusion coefficient was also applied to Ponderosa pine, but may
overestimate the characteristic times for this wood species. Theoretical calculations of the
relative tangential conductances for various pine species have yielded 0.03 for Tamarack
(Behr et al., 1953) and Western white pine (Burr and Stamm, 1947).
The characteristic times for physically hindered diffusion through wood are reported in
Table 6.5. These values were calculated using the wood tangential diffusion coefficient and
the aqueous diffusivities reported in Table 6.4. The estimated characteristic times were of the
same order of magnitude as the measured t 1/2times. The tl/ 2 value for o-xylene uptake by
Douglas fir was greater than estimated by the hindered diffusion model.
The wood diffusion coefficient was recalculated accounting for sorption to cell wall
tissue. In this case, the relative conductances through the pit membranes and cell walls was
decreased by a factor of 10, an order of magnitude representation of the wood retardation
factors. (The overall conductance was insensitive to the factor of 3 variation in retardation
factors between benzene, toluene and o-xylene.) The overall conductance was about one third
of the value without tissue retardation, yielding characteristic times greater than the measured
values (Table 6.5).
The wood sorption kinetics for benzene, toluene and o-xylene had t,/2 times that fell
between estimates with solely hindered diffusion and hindered diffusion with cell tissue
retardation. From our isotherm experiments, these compounds are known to sorb to the wood
cell wall tissue. It is not known how the tendency to sorb affects the diffusional mechanism
for these sorbates. For example, if surface diffusion is important, times may be faster than
required for retarded cell tissue diffusion which assumed free liquid diffusion. The low heats
of sorption and the noncompetitiveness of sorption to wood (Severtson and Banerjee, 1996)
suggest an absorptive partitioning to wood; however, there is adsorptive compound uptake by
wood at high sorbate concentrations. Sorption isotherms for stearic acid and phenol have
been used to estimate internal surface area of sugar pine (Stamm and Millet, 1941). If wood
sorption
203
Table 6.5. Estimated characteristic mass transfer times (min) for hindered and retarded
diffusion.
Diffusion Model
Benzene
Toluene
o-Xylene
100
110
130
Stamm's model
with retardation
(Figure 6.11 b)
260
300
330
Homogeneous
porous sorbent with
retarded diffusion
60
80
330
Stamm's model
(Figure 6.11a)
204
I
..1. .^ ,_...r.xi.ru.rx-rusl~-------.--.~-~-----I -----~-----~
l-----~L
i~-
~--r-~PI~
I--L~-
---IXr -----~~'~~-
1
I
was an adsorptive process, surface diffusion may be an important internal mass transfer
process.
CharacteristicDiffusion Times - Homogeneous Retarded Diffusion
The second model for sorbate uptake by wood assumed a homogeneous porous sorbent
with retarded diffusion. The sorbate aqueous diffusivities were divided by the Douglas fir
retardation factors to give the characteristic diffusion coefficients used in Eq. 7. The
characteristic times for retarded diffusion were of the same order of magnitude as the
measured values with the exception of underestimating the Douglas fir-o-xylene t 1/2time.
The applicability of a physically hindered diffusion model or a homogeneous retarded
diffusion model for wood diffusion kinetics can not be concluded from these experimental
results. For the sorbates used, the characteristic times calculated with both of these models
were the same, and in general agreement with the experimental observations. The sorption
kinetics of a more hydrophobic compound may suggest which model better describes the
uptake of sorbing compounds by wood. For example, a compound such as
pentamethylbenzene, with a Kow of 1046 (Hansch et al., 1995), has a retardation factor of 165
(assuming the Ponderosa pine lignin-octanol LFER, f1,gnmn
of 0.3, and rsw, of 0.6 g/mL).
Stamm's diffusion model with cell wall retardation reaches an asymptotic relative conductance
of 0.012 for compounds of this hydrophobicity. The corresponding characteristic diffusion
time for pentamethylbenzene (370 min) is much less than the characteristic time of 750 min
for retarded diffusion of this compound. If the measured tl/2 value for pentamethylbenzene
were closer to the values obtained for BTX, we would conclude that diffusion into wood is
not adequately represented by assuming wood is a homogeneous porous sorbent. Further
investigation into the mechanism for diffusion of sorbing compounds into wood was beyond
the scope of this investigation.
Environmental Relevance
The results of our equilibrium sorption study have implications with respect to the
transport of groundwater contaminants at industrial sites with wood present in the aquifer
solids. At these sites, plume movement of organic compounds would be underestimated using
an organic carbon-based retardation model. The fraction of lignin in the solids must be
205
II
--
quantified and a lignin LFER used to predict the wood partition coefficient. The applicability
of our softwood lignin-octanol free energy relationship to hardwoods was not investigated in
this study, but, as noted, softwoods may be the dominant species of woods at industrial sites
due to their prevalence in construction lumber. Wood and paper products are also found in
landfills. From 1 to 4% by weight of municipal refuse is wood, although this percentage is
greater for landfills that accept demolition wastes and paper products compose 30 to 60% of
refuse by weight (Niessen, 1977). Estimates of leachate transport through these fill solids
must account for the high fraction of cellulose carbon which exhibits almost no sorption of
organic contaminants.
The timescale for kinetic exchange of contaminants between wood and groundwater
may be fast relative to other environmental processes. The actual timescale is a function of
the average wood particle size, but for cm-sized wood chips, sorption kinetics were much
faster than slow diffusion from natural organic matter (Ball and Roberts, 1991; Steinberg et
al., 1987). Thus, diffusive exchange from wood would not be rate-limiting relative to the
timescales of pump-and-treat remediation technologies.
206
^.
---.-*~- -- 111 1
-l--.~1111~-~^l
1IIII_ -Ill~-C 1~- ..1-~-
References
Ball, W. P.; Roberts, P. V. (1991). "Long-term sorption of halogenated organic chemicals by
aquifer materials. Part 2. Intraparticle diffusion." Environmental Science and
Technology 25: 1237-1249.
Barton, A. F. M. (1975). "Solubility parameters." Chemical Reviews 75: 731-753.
Barton, A. F. M. (1983). CRC Handbook of Solubility Parametersand Other Cohesion
Parameters.Boca Raton, FL, CRC Press, Inc.
Behr, E. A.; Briggs, D. R.; Kaufert, F. H. (1953). "Diffusion of dissolved materials through
wood." Journalof Physical Chemistry 57: 476-480.
Bodig, J. (1982). "Moisture Effect on Structural Use of Wood." In Structural Use of Wood in
Adverse Environments. R. W. Meyer ,R. M. Kellogg, Eds.. New York, Van Nostrand
Reinhold Co.
Burr, H. K.; Stamm, A. J. (1947). "Diffusion in wood." Journal of Physical and Colliod
Chemistry 51: 240-261.
Cady, L. C.; Williams, J. W. (1935). "Molecular diffusion into wood." Journal of Physical
Chemistry 39: 88-103.
Chin, Y.-P.; Aiken, G. R.; Danielson, K. M. (1997). "Binding of pyrene to aquatic and
commercial humic substances: The role of molecular weight and aromaticity."
Environmental Science and Technology 31: 1630-1635.
Chiou, C. T.; Peters, L. J.; Freed, V. H. (1979). "A physical concept of soil-water equilibria
for nonionic compounds." Science 206: 831-932.
Crank (1979). The Mathematics of Diffusion. Oxford, Clarendon Press.
Dellicoli, H. T. (1977). "Controlled Release of Pesticides from Kraft Lignin Carriers." In
ControlledRelease Pesticides. H. B. Scher, Ed. Washington, DC, American Chemical
Society.
Fengel, D. (1971). "The distribution of aqueous solutions within wood." Journal of Polymer
Science, Part C 36: 141-152.
207
_
dlD
Garbarini, D. R.; Lion, L. W. (1986). "Influence of the nature of soil organics on the sorption
of toluene and trichloroethylene." Environmental Science and Technology 20:
1263-1269.
Genco, J. M. (1996). "Pulp." In Encyclopedia of Chemical Technology. J. I. Kroschwitz, Ed.
New York, Wiley.
Hansch, C.; Leo, A.; Hoekman, D. (1995). Exploring QSAR: Hydrophobic, Electronic and
Steric Constants. Washington, DC, American Chemical Society.
Higuchi, T. (1980). "Lignin Structure and Morphological Distribution in Plant Cell Walls." In
Lignin Biodegradation:Microbiology, Chemistry and PotentialApplications. T. K.
Kirk, T. Higuchi ,H.-M. Chang, Eds.. Boca Raton, FL, CRC Press. 1.
Karickhoff, S. W.; Brown, D. S.; Scott, T. A. (1979). "Sorption of hydrophobic pollutants on
natural sediments." Water Research 13: 241-248.
Kopinke, F.-D.; Porschmann, J.; Stottmeister, U. (1995). "Sorption of organic pollutants on
anthropogenic humic matter." Environmental Science and Technology 29: 941-950.
Kumada, K. (1987). Chemistry of Soil Organic Matter. Tokyo, Japan Scientific Societies.
Luthy, R. G.; Dzombak, D. A.; Peters, C. A.; Roy, S. B.; Ramaswami, A.; Nakles, D. V.;
Nott, B. R. (1994). "Remediating tar-contaminated soils at manufactured gas plant
sites." Environmental Science and Technology 28: 266A-276A.
Mackay, D.; Shiu, W. Y.; Ma, K. C. (1992). IllustratedHandbook of Physical-Chemical
Propertiesand Environmental Fate of Organic Chemicals. Boca Raton, FL, Lewis
Publishers.
Niessen, W. R. (1977). "Properties of Waste Materials." In Handbook of Solid Waste
Management. D. G. Wilson, Ed. New York, Van Nostrand Reinhold.
Parham, R. A.; Gray, R. L. (1984). "Formation and Structure of Wood." In The Chemistry of
Solid Wood. R. Rowell, Ed. Washington, DC, The American Chemical Society. 207:
3-56.
Pearl, I. A. (1967). The Chemistry of Lignin. New York, Marcel Dekker.
Petterson, R. C. (1984). "The Chemical Composition of Wood." In The Chemistry of Solid
Wood. R. Rowell, Ed. Washington, DC, The American Chemical Society. 207: 59-126.
208
~_l_--_
- rr~-----i--*l^-r*r~--rrx~-.
rr
..il~---_El-----^
----r~-L--^-
~-- -XI---LI*-r---~_
_Ill_----L-~ ~-~~~--~--~-
I
~~-1~-----1
I
__
~--~---~--~s
Rael, J.; Shelton, S.; Dayaye, R. (1995). "Permeable barriers to remove benzene: Candidate
media evaluation." Journal of EnvironmentalEngineering 121: 411-415.
Rutherford, D. W.; Chiou, C. T.; Kile, D. E. (1992). "Influence of soil organic matter
composition on the partition of organic compounds." Environmental Science and
Technology 26: 336-340.
Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M. (1993). Environmental Organic
Chemistry. New York, NY, John Wiley & Sons, Inc.
Schwarzenbach, R. P.; Westall, J. (1981). "Transport of nonpolar organic compounds from
surface water to groundwater: Laboratory sorption studies." Environmental Science
and Technology 15: 1360-1367.
Severtson, S. J.; Banerjee, S.(1996). "Sorption of chlorophenols to wood pulp."
Environmental Science and Technology 30: 1961-1969.
Siau, J. F. (1984). Transport Processes in Wood. New York, Springer-Verlag.
Stamm, A. J. (1964). Wood and Cellulose Science. New York, The Ronald Press Company.
Stamm, A. J.; Millet, M. A. (1941). "The internal surface area of cellulosic materials."
Journalof Physical Chemistry 45: 43-54.
Steinberg, S. M.; Pignatello, J.J.; Sawhney, B. L. (1987). "Persistence of 1,2-dibromoethane
in soils: Entrapment in intraparticle micorpores." Environmental Science and
Technology 21(12): 1201-1208.
Thompson, N. S.(1996). "Hemicellulose." In Encyclopedia of Chemical Technology. J.I.
Kroschwitz, Ed. New York, Wiley.
Wegner, T. H.; Baker, A. J.; Bendtsen, B. A.; Brenden, J.J.; Eslyn, W. E.; Harris, J.F.;
Howard, J.L.; Miller, R. B.; Petersen, R. C.; Rowe, J.W.; Rowell, R. M.; Simpson,
W. T.; Zinkel, D. F. (1992). "Wood." In Kirk-Other Encyclopedia of Chemical
Technology. J.I. Kroschwitz ,M. Howe-Grant, Eds.. New York, John Wiley and Sons.
Wu, S.-C.; Gschwend, P. M. (1988). "Numerical modeling of sorption kinetics of organic
compounds to soil and sediment particles." Water Resources Research 24: 1373-1383.
209
-
-- -~
Xing, B.; McGill, W. B.; Dudas, M. J. (1994). "Cross-correlation of polarity curves to predict
partition coefficients of nonionic organic contaminants." Environmental Science and
Technology 28: 1929-1933.
210
.~~~,~_~,~~^,.,...,,,~- i~r^-.a----- ^~~LIIP-~---Il-~ .^i
I
Chapter 7.
SUMMARY OF RESULTS AND
FUTURE STUDY OF INDUSTRIAL SITES
211
~
I
Introduction
This thesis was undertaken with the broad goal of understanding the unique
characteristics of contaminated sites with a history of industrial activity that has resulted in
significant anthropogenic reworking of the subsurface solids. A former manufactured gas
plant with a 150 year history of industrial operation was investigated as a representative
example of such sites. Characterization of the unique properties of the subsurface solids and
groundwater transport processes proceeded on several fronts involving both laboratory and
field work. In this chapter, the reverse process is followed and the various lines of
investigation are drawn together to predict the transport of contaminants at this coal tar site.
These calculations demonstrate that the anthropogenic fill solids have transport properties that
vary significantly from natural aquifers. The scope of the discussion is then broadened to
discuss areas of further investigation which would enhance our understanding of groundwater
transport and reaction processes at industrial sites. A general approach for remedial
investigations at these sites is also outlined.
Summary of Results
The groundwater transport of organic contaminants is a function of both the mobile
contaminant load and individual compound interactions with the solid phase. At Site YYZ, a
former manufactured gas plant with subsurface coal tar contamination, the source of aromatic
hydrocarbons in the groundwater was found to be equilibrium coal tar dissolution. Evidence
of colloid-enhanced groundwater solubilization and biodegradation of compounds were found
by quantifying groundwater concentrations of individual compounds from this multicomponent source. The subsurface solids, a mixture of manufactured gas production wastes
and anthropogenic fill, contained various organic carbon-containing materials in addition to
natural organic matter. These fill solids exhibited solid-water partition coefficients that were
unique from natural organic matter.
The solid-water partitioning of monoaromatic compounds to wood was investigated
further. A linear free energy relationship was developed to enable prediction of wood
partition coefficients from octanol water partition coefficients. The sorption kinetics of wood
212
.---u-L---- ---r--.-.r
- --ILLIIC-II~---.-l~
rr~-_r-...-Llr.xr^r.i~--r~_ra~^---^--
~II~------l--~-__I
.~ .~~----i---~-I CI1111~-~. _~I~C-
_I~~_I_~_I
1
particles similar in size to those found at Site YYZ were fast relative to groundwater
advection.
Application of Results to Transport Calculations at Site YYZ
The effect of colloid-enhanced solubility and the presence of various organic carboncontaining fill solids on groundwater transport at Site YYZ were quantified by calculating
retardation factors for these two cases. The calculated retardation factors were compared to
estimates of retardation with no colloid-association of contaminants and natural organic
matter-dominated solid-water partitioning. These calculations are a hypothetical
demonstration of the importance of accurately characterizing groundwater phase transfer and
reaction processes at other industrial sites. As discussed in Chapter 2, the groundwater at
Site YYZ is saturated with respect to coal tar constituents. Thus, predictions of transport,
such as the travel time for a contaminant from to impact a sensitive receptor (e.g., a drinking
water well), are not applicable to the anthropogenic fill at this coal tar site. The following
retardation factor calculations assume groundwater transport away from the source into a
water-bearing unit not already impacted by contaminants.
The influence of colloids on the rate at which a breakthrough front of hydrophobic
aromatic hydrocarbons travels relative to the rate with no colloidal transport depends upon the
model for contaminant exchange between the colloid, solid and aqueous phases and the model
for colloid exchange between the mobile and immobile phases (Corapcioglu and Jiang, 1993).
Under the assumption of linear equilibrium partitioning of the contaminant to the colloid and
solid phases, and linear equilibrium partitioning of the colloids to the solid phase, and a
constant groundwater colloid concentration, the retardation factor is given as (Corapcioglu and
Jiang, 1993):
R =1 +
r K
sw d(1)
1 + (COLL) KOLL
where r,w (g/mL) is the solid-to-water ratio; Kd (mL/g) is the solid-water partition coefficient;
(COLL) (g/mL) is the concentration of colloids in the groundwater, and KCOLL (mL/g) is the
colloid-water partition coefficient. This equation is applicable to suspended organic matter
213
colloids which have a lesser tendency to partition to the immobile solid phase than do
polycyclic aromatic hydrocarbons (Magee et al., 1991). For PAHs, the product of the solidto-water ratio and the partition coefficient is generally much greater than 1 so the retardation
factor for these compounds in colloid-containing groundwater is reduced by a factor of
(1 + (COLL)KcoLL) relative to a system with no colloids. This factor is equal to the
enhancement factor calculated by taking the ratio of enhanced solubilities (observed
groundwater concentrations, Chapter 2) to compound solubilities with no colloids present
(measured tar-water equilibrium, Chapter 2). The enhancement factors for benzo(a)pyrene at
well W40M ranged from 4 to 16. In a hypothetical groundwater aquifer with similar colloid
characteristics, the retardation factor for benzo(a)pyrene would be reduced by a factor of 4 to
16 relative to predictions assuming no colloids. For a typical natural aquifer (f,, =
10 -3
g/g),
the benzo(a)pyrene retardation factor would decrease from about 1000 to about 100 with this
magnitude of colloid-enhanced solubility. This colloid-impacted retardation factor for
benzo(a)pyrene is still high, but if the industrial site that was the source of this groundwater
contamination had been in operation for over 100 years, sufficient time may have elapsed for
benzo(a)pyrene to have been transported to sensitive receptors, even with a retardation factor
of 100.
Note that the retardation factor was used here as a convenient comparator for assessing
colloid-facilitated groundwater contaminant transport. The retardation factor does not account
for the increased flux (over the only dissolved case) of contaminants away from the source
facilitated by colloid-association of a population of contaminant molecules. To predict the
true impact of colloidal transport on contaminant mobility a set of coupled transport equations
are required - one transport equation for the dissolved contaminant and one transport equation
for the colloid phase (Corapcioglu and Jiang, 1993).
Naphthalene retardation factors were calculated for particular depths in the fill solids at
which only one type of organic carbon-containing material was found. For the purposes of
demonstration, this calculation did not account for the presence of nonaqueous phase liquids
(NAPL) in the solids. At all depths, solid-water partitioning was dominated by tar-water
partitioning. In Chapter 5, natural organic matter-based partition coefficient calculations were
214
I~~~u;il~r.-..li.l~,.;..
i-rla
- ^---~-rl---
~u.rmr~-xr~----IF~lll-^'IX1~
--p--
. ____ ___ ..--~II--~~--
^~
_
-111
--
_~___l--C~-
Table 7.1. Naphthalene retardation factors as a function of depth at Site YYZ.
Depth
Sorbent
foc*
Rt
(Kd
foKoc)
R
(K d =
fiKi)
3.0 m
wood chips
0.04
170
36
3.7 m
coke
0.54
2300
1
6.7 m
natural solids
0.001
5
5
'Fraction organic carbon calculated by subtracting the nonaqueous phase liquid content
determined by residue on evaporation of solvent extraction, assuming 80% carbon content,
from the total organic carbon at the depth (values reported in Figure 5.1, Chapter 5).
t R = 1 + rswKrad
where rsw (g/mL) is the solid-to-water ratio, here assumed 5 g/mL
and Ktrad = focKo (mL/g)
where foc (g/g) is the fraction organic carbon and Koc = 850 mL/g for naphthalene
(Chapter 5).
R = 1 + rswKd
where Kd = f, K, (mL/g)
and f,,gn, = 0.3 fo, Kignm = 580 (mL/g), and Kcoke = 0 (mL/g) (Chapter 5).
215
shown to underestimate the solid-water partition coefficients, and hence retardation factors, in
NAPL-containing solids. Here, natural organic matter-based partition coefficients are shown
to overpredict retardation in aquifer solids composed of other organic-carbon containing
materials (Table 7.1). Retardation factors were underestimated by factors of 5 (wood) and
2300 (coke). Therefore, sorbent-specific partition coefficients are required to estimate
contaminant transport through organic carbon-containing aquifer solids.
Areas of Further Investigation
Further investigation into understanding groundwater fate processes is required in order
to predict which mechanisms would be operative at other industrial sites as a result of the
history of industrial processes at those sites. An important fate process that requires further
investigation is the conditions under which organic colloids are present in groundwater.
Organic colloids in contaminant plumes show an increased capacity to sorb hydrophobic
contaminants over organic colloids in pristine environments (Hawley, 1996; Backhus and
Gschwend, 1990), but the source and composition of these colloids are not known. It has
been suggested that organic colloids are contaminant molecules that have been partially
metabolized by microbes (Hawley, 1996). In fill solids at industrial sites, organic colloids
may also be generated by microbial degradation of the fill materials, such as wood. A
understanding of the sources of organic colloids in contaminant plumes may also help to
explain the temporal (Chapter 2, Backus (1990)) and spatial (Chapter 2) variability in colloid
presence in groundwater.
The occurrence of biodegration at industrial sites is a second fate process that is not
well understood. Microbes have been shown to degrade a variety of different organic
contaminants with many different terminal electron acceptors. Industrial sites have great
potential for increased bioattenuation over natural aquifers because of the wide variety of fill
materials and wastes that may have been disposed. For example, at Site YYZ buried pipes
may have been a source for methane, iron-containing minerals were buried in the subsurface
and the groundwater contained high levels of sulphate. With a better understanding of the
primary mechanisms of microbial degradation, predictions of the potential for contaminant
216
l--L-r^-X~----^----~-r*lr*pl --rrYII
~.~n~~r~-c
^
,...I-x~ ~xu.~rr-.
--I--~~
I_-...~-^-.-
~p_
bioattenuation at an industrial site may be made with knowledge of the industrial processes
and the possible waste products that could be buried on site.
There is a need to incorporate multiple-component contaminant sources in transport
models. The relative distributions of compounds with varied physical-chemical properties in
space (site investigation monitoring wells) and time (pump tests) can give information about
potential rate limitations or natural attenuation processes before remediation technologies are
implemented. Remedial approaches can then be designed to account for these processes.
Better methods are required to quantify the various organic-carbon-containing fill
solids in anthropogenic fill materials. Selective digestions were proposed as a method to
quantify various sorbent materials in a mixture. The presence of nonaqueous phase liquids in
aquifer are not well quantified. NAPLs that are less dense than water can be quantified by
absorption with a foam plug after release from the solids in an aqueous medium (Cary et al.,
1991). No similar method exists for the quantification of nonaqueous phase liquid mixtures,
such as coal tar, that are denser than water. These NAPLs are quantified by bulk solvent
extraction that also includes sorbed compounds. A method that tries to assess the presence of
NAPL by subsequently calculating whether the aqueous solubility of a compound would have
been exceeded in the pore water requires prior knowledge of the NAPL composition and may
mistake small concentrations of NAPL as no NAPL (Feenstra et al., 1991). These small
concentrations may still be abundant enough to limit pump-and-treat based remediation
(MacKay et al., 1996).
General Approach to Remedial Investigations of Contaminated Sites
with a History of Industrial Activity
The groundwater contaminants and the subsurface solids at a particular industrial site
are unique to the history and manufacture of operations that occurred at the site. A general
guideline to remedial investigation at an industrial site must begin with a review of the
history of manufacture to know possible reagents and starting materials, products and wastes
that were used or generated on the site. These are the most likely materials to compose onsite anthropogenic fill and groundwater contaminants. While the actual investigation and
remediation approaches will be unique to the property, it is necessary to gather concentration
217
data for compounds with varied physical-chemical properties (e.g., octanol water partition
coefficient, Henry's law constant, and approriate others) to help understand groundwater fate
processes. Individual compound analyses require better analytical precision than bulk
measures of contamination (e.g., total petroleum hydrocarbons, total volatile organic
compounds); however, this added expense at strategic sample locations (in space or time) may
help to save remediation costs. For example, remediation costs may be lowered by
identifying rate-limitations to clean-up before remediation systems are designed. At sites
where nonaqueous phase liquids are present, calculations of the expected equilibrium aqueous
phase concentrations should be made using Raoult's Law. These values will provide a
comparison case for groundwater monitoring to gather information about dominant
groundwater processes and sampling artifacts. Finally, the future clean-up of other industrial
sites will benefit from the dissemination of information about unique obstacles to remediation
and innovative approaches used at industrial sites. There are many classes of sites which
have similar groundwater contaminants and waste products (e.g., coal tar sites), and thus
publication of case studies in peer-reviewed literature will increase the body of knowledge
about these industrial sites and aid the design of investigative and remediation approaches at
other similar sites.
218
.rl.--- P~---l-i------~li-~-Lnrlll
^
L^----I~--L-C--Y-~-
~---^IIICI..~_1_1_~_II li--I~-
------
-I
~tP---^-~
I
_-----------~
~--~--~-
References
Backhus, D. A. (1990). Colloids in Groundwater: Laboratory and Field Studies of Their
Influences on Hydrophobic Organic Contaminants.Ph.D. Thesis, Massachusetts
Institute of Technology.
Backhus, D. A.; Gschwend, P. M. (1990). "Fluorescent polycyclic aromatic hydrocarbons as
probes for studying the impact of colloids on pollutant transport in groundwater."
Environmental Science and Technology 24: 1214-1223.
Cary, J. W.; McBride, J. F.; Simmons, C. S. (1991). "Assay of organic liquid contents in
predominantly water-wet unconsolidated porous media." Journal of Contaminant
Hydrology 8: 135-142.
Corapcioglu, M. Y.; Jiang, S. (1993). "Colloid-facilitated groundwater contaminant transport."
Water Resources Research 29: 2215-2226.
Feenstra, S.; Mackay, D. M.; Cherry, J. A. (1991). "A method for assessing residual NAPL
based on organic chemical concentrations in soil samples." Ground Water Monitoring
Review 11(Spring): 128-136.
Hawley, C. M. (1996). A Field and Laboratory Study of the Mechanisms of Facilitated
Transport of Hydrophobic Organic Contaminants.M.S. Thesis, University of
Colorado.
MacKay, A. A.; Chin, Y.-P.; MacFarlane, J. K.; Gschwend, P. M. (1996). "Laboratory
assessment of BTEX soil flushing." Environmental Science and Technology 30:
3223-3231.
Magee, B. R.; Lion, L. W.; Lemley, A. T. (1991). "Transport of dissolved organic
macromolecules and their effect on the transport of phenanthrene in porous media."
Environmental Science and Technology 25: 323-331.
219
----
^--
Appendix A.
AQUEOUS SOLUBILITY OF AROMATIC HYDROCARBONS
IN EQUILIBRIUM WITH COAL TAR
220
^,~,~
~_~,.~...~.,.._....~..~~~.~
-~--------- ~
~I___,
..--- ---------------------- ---- - ---- -----------I----
~-~----
-
I
Introduction
The expected equilibrium groundwater concentrations of organic contaminants are
often of interest at sites with nonaqueous phase liquids (NAPLs) present. Contaminants will
dissolve from the NAPL source until individual compound fugacities are equal in the
groundwater and the nonaqueous phase. The equilibrium aqueous compound concentrations
can be predicted with knowledge of the NAPL composition and theoretical expressions for the
ratio of phase fugacities (Schwarzenbach et al., 1993; Prausnitz et al., 1986). For multicomponent nonaqueous phase mixtures, dissolution of individual compounds is expected to
obey Raoult's Law:
Cw = YNXCsat(L)
(1)
where C, (mg/L) is the equilibrium aqueous concentration, yN is the activity coefficient in the
NAPL phase, XN is the compound mole fraction in the NAPL phase, and Csat (L) (mg/L) is
the saturated aqueous phase solubility of the pure liquid compound.
Raoult's Law for mixtures of organic compounds is modified by the inclusion of a
nonaqueous phase activity coefficient, yN. This variable accounts for differences in molecular
size, shape and intermolecular forces between the solute compound and the solvent NAPL
phase. The magnitude of these solution nonidealities cannot easily be predicted, although
empirical expressions for activity coefficients have been developed (Prausnitz et al., 1986).
Generally, organic liquid phases in the environment are mixtures of similar compounds (e.g.,
chlorinated solvents, aliphatic hydrocarbons, etc.) and thus ideal solubility of individual
components in the NAPL phase is assumed (i.e., y, = 1).
The equilibrium aqueous phase concentration of a compound, C,, is proportional to its
saturated aqueous solubility as a pure liquid compound. For compounds that are liquids at
environmental temperatures, Csat (L) is the compound aqueous solubility. For compounds
which are solids, Csat (L) is the aqueous solubility of the subcooled liquid. Solids which are
dissolved in nonaqueous phase mixtures have freedom to rotate as if they were in a liquid
state at the system temperature. A subcooled liquid is a theoretical thermodynamic phase.
Pure compounds cannot exist as subcooled liquids, but the subcooled liquid solubility of a
221
I
%-
compound can be estimated from the aqueous solubility of the pure solid, corrected for the
enthalpic cost of melting:
Csa(L) =Csat,(s) exp
AH
RT
MTmT - 1
(2)
where Csat (s) (mg/L) is the pure solid aqueous solubility, AH (J/mol) is the enthalpy of
melting, R (J/mol-K) is the gas constant, Tm (K) is the compound melting temperature, and
T (K) is the system temperature. Eq. 2 only considers the energy cost of the solid-to-liquid
phase change and neglects heat capacity terms. These additional terms which account for the
enthalpy of heating the solid and cooling the liquid, before and after the phase change, are
usually insignificant (Prausnitz et al., 1986). When they are not, they can lead to large errors
in Csat (L) calculations because they change the exponential term (Mukherji et al., 1997).
Raoult's Law (y, = 1) calculations of aqueous compound concentrations in the
presence of complex NAPL mixtures have shown good agreement with aqueous
concentrations measured in batch equilibrations. In a study of seven coal tars, measured
concentrations of aromatic hydrocarbons varied by less than a factor of 2 from predicted
values (Lee et al., 1992). Coal tars are 40 to 100% aromatic compounds by weight (Nelson
et al., 1996), thus NAPL activity coefficients for aromatic hydrocarbons would likely be near
1. Aqueous equilibrium with dissimilar NAPL mixtures could also be predicted with y, = 1
for alkylated benzenes dissolving from crude oil (Eganhouse et al., 1996). Nonideal solution
behavior might be expected for aromatic hydrocarbons dissolved in this aliphatic nonaqueous
phase. For example, the activity coefficient for benzene in hexane is 2 (Schwarzenbach et al.,
1993).) These batch equilibrations of water and nonaqueous phase liquids suggest that
Raoult's Law can be used to predict equilibrium groundwater concentrations at sites with
complex nonaqueous phase liquids present, if the NAPL composition is known.
The source of groundwater contamination at Site YYZ was equilibrium coal tar
dissolution (Chapter 2). To investigate the groundwater fate of aromatic hydrocarbons at this
site, compound concentrations were compared to measured aqueous concentrations from a tar-
222
,r*il
-----~--~,~-~rm~----- ~
-p-'^-~^"~L--L--LI~~-li~ -Y-~LIII
II~I^I__^I.I~_~~
1~1-.-I~C~--I
._
LIX-C----.~--..-IIII^Y~PI----- -_II
~----~-- -~~~1~~1_1
water batch equilibration. In this appendix, the measured aqueous concentrations are
compared to Raoult's Law calculations of equilibrium aqueous concentrations.
Methods
Measurements of aqueous aromatic hydrocarbon concentrations in equilibrium with
Site YYZ coal tar were made according to the methods detailed in Chapter 2 for tar-water
equilibration.
Calculated aqueous concentrations of aromatic hydrocarbons in equilibrium with
Site YYZ coal tar were made with Eq. 1, assuming ideal solubility. Compound mole
fractions were calculated from analysis of the tar phase (Chapter 2):
t10
06
XN = Car
Ptar
MW
(3)
or
MW
where Ctr (mg/L) is the compound concentration in the tar, P,, (g/mL) is the measured tar
density, 10-6 (L-g/mL-mg) is a unit conversion, and MW,,r and MW,are the molecular weights
of the tar and the compound, respectively. Ca values are reported in Table 2.2 (Chapter 2).
The tar density was measured to be 1 g/mL. The tar molecular weight was assumed to be
160 g/mol (EA Engineering Science and Technology Inc, 1993). This value was
representative of tars isolated from other monitoring wells in the vicinity of the MIT
monitoring wells and is consistent with other liquid tars (Lee et al., 1992). Subcooled liquid
solubilities were taken from Miller et al. (1985). Only solid solubilities were reported for
2-methylnaphthalene, chrysene, and benzo(ghi)perylene. Subcooled liquid solubilities were
calculated for these compounds with an enthalpy of melting equal to 56.5 J/mol-K
(Schwarzenbach et al., 1993).
223
x
Tm
Results and Discussion
Calculated equilibrium concentrations were compared to measured aqueous
concentrations of coal-tar equilibrated aromatic hydrocarbons (Figure A. 1). Error bars
representative of the analysis methods and concentration calculations are also included. The
measurement error for the volatile compounds was taken as the variability in internal purge
standards (10%, Chapter 2). Polycyclic aromatic hydrocarbon concentrations had a
measurement error of 28% determined from internal standard recoveries in the solvent
extraction method.
There were a number of sources of error in the Raoult's Law calculation of compound
aqueous concentrations. First, the tar molecular weight was chosen from wells in the vicinity
of the MIT monitoring wells because these values were thought to be representative of tar in
the study area. Tar molecular weights across the site varied with an average value of
195 ± 45 g/mol (n = 7) (EA Engineering Science and Technology Inc, 1993). Thus, tar
molecular weight values could be in error by 23%.
Subcooled liquid solubilities are the major source of error in calculations with Eq. 1.
First, as detailed in Eq. 2, good measurements of the pure compound solubilities are required
to calculate Csat (L). A wide range of solubility values have been reported for low solubility
polycyclic aromatic hydrocarbons (PAHs) (Mackay et al., 1992). Mukherji et al. (1997)
estimated a range of ± 25% in reported solubility values for PAHs. Secondly, a major
assumption in calculating subcooled solubilities is the assumption of a constant entropy (and
hence enthalpy) of melting for these rigid molecules (Schwarzenbach et al., 1993). Entropies
of melting for PAHs vary from 59 to 38 J/mol-K while a constant value of 56.5 J/mol-K is
assumed (Miller et al., 1985). For pyrene (38J/molK, Wauchope and Getzen, 1972) Csat (L)
values are overestimated by a factor of 3 by assuming a constant entropy value for all PAHs.
The errors in tar molecular weight and pure compound solubilities were propagated to
give an error of 40% in the calculated Cw values. Errors due to phase change calculations
(AHm, omission of heat capacity terms) were omitted because of the good agreement of other
experimental data with Raoult's Law calculations using Eq. 2 with AHm = 56.5 J/mol-K
224
- ,~.
-
x
Tm.
_~...~a~
~-----C^r~-..~~..,_~,~~~~.
----I-~------"1-Il^
le+2
le+1
le+O
le-1
le-2
le-3
le-4
le-5
'_.
le-5
le-4
le-3
le-2
le-1
le+0
le+l
le+2
Measured Aqueous Concentration (mg/L)
Figure A.1. Comparison of calculated and measured aqueous mono- and polycyclic
aromatic hydrocarbon concentrations in equilibrium with W40M coal tar. Compound
identities are abbreviated in Table 2.2. Error bar calculations are detailed in the text.
Where not visible, error bars are smaller than symbols.
225
II
Within the measurement and estimation errors, measured aqueous concentrations were
within a factor of 2 of Raoult's Law calculations for Site YYZ coal tar. Pyrene,
benzo(a)pyrene and benzo(ghi)perylene showed the greatest deviation from the ideal solubility
line. If the source of disagreement between the calculated and observed concentrations was
an incorrect tar molecular weight value used in Eq. 3, all of the data points would have fallen
above the y, = 1 line. (The value of 160 g/mol was less than the site average suggesting that
if this value was not representative of W40M tar, the true molecular weight of this tar would
have been greater.) Data points fall above and below the ideal solubility line so poor
estimates of the tar molecular weight likely do not explain differences between calculated and
measured solubilities.
The differences in calculated and measured aqueous concentrations likely result from
inexact estimates of compound subcooled liquid solubilities. Calculations of pyrene aqueous
concentrations could be improved by use of its measured enthalpy of melting. As noted
earlier, the subcooled liquid solubility used to calculate Cw may have overestimated pyrene
solubility by factor of 3. It is difficult to measure accurate saturated aqueous concentrations
of benzo(a)pyrene and benzo(ghi)perylene because of the low solubilities of these compounds.
Deviations from ideal solubility (yN = 1 line) may result from low values of Csat (s) or
measurement error in this tar-water equilibration.
The measured concentrations of aromatic hydrocarbons were compared to observed
groundwater concentrations, rather than comparing Raoult's Law estimates because of the
uncertainties in Csat (L) values. Also, any systematic errors in our analytical method would
cancel one another when comparing groundwater observations to measured tar-water
equilibrium concentrations.
In conclusion, Site YYZ coal tar is another example of a complex environmental
organic mixture for which Raoult's Law predicts the equilibrium aqueous concentrations of its
component compounds. Given the sources of uncertainty in the calculation of subcooled
liquid solubilities and the difficulty in measuring aqueous concentrations of low solubility
organic compounds, agreement of Raoult's Law calculations within a factor of 2 of aqueous
concentrations in batch equilibrations is reasonable.
226
~~..~_~I~~-~..~l....~.r.~~. .....- ..,.~.,~_r~----------I---1
---
~~I~~'P-~-*-"III~.P-.
1 L-~1
~_pl-~-^-il-----~_ll_^_~
P___
_slll~^~llll~--.IX~~~ -~-
--
I
References
EA Engineering Science and Technology Inc (1993). Site CharacterizationReport for the
BG&E Spring Gardens Facility.
Eganhouse, R. P.; Dorsey, T. F.; Phinney, C. S.; Westcott, A. M. (1996). "Processes affecting
the fate of monoaromatic hydrocarbons in an aquifer contaminated by crude oil."
Environmental Science and Technology 30: 3304-3312.
Lee, L. S.; Rao, P. S. C.; Okuda, I. (1992). "Equilibrium partitioning of polycyclic aromatic
hydrocarbons from coal tar into water." Environmental Science and Technology 26:
2110-2115.
Mackay, D.; Shiu, W. Y.; Ma, K. C. (1992). IllustratedHandbook of Physical-Chemical
Propertiesand Environmental Fate of Organic Chemicals. Boca Raton, FL, Lewis
Publishers.
Miller, M. M.; Wasik, S. P.; Huang, G. L.; Shiu, W. Y.; Mackay, D. (1985). "Relationships
between octanol-water partition coefficient and aqueous solubility." Environmental
Science and Technology 19: 522-529.
Mukherji, S.; Peters, C. A.; Jr, W. J. W. (1997). "Mass transfer of polynuclear aromatic
hydrocarbons from complex DNAPL mixtures." Environmental Science and
Technology 31: 416-423.
Nelson, E. C.; Ghoshal, S.; Edwards, J. C.; Marsh, G. X.; Luthy, R. G. (1996). "Chemical
characterization of coal tar-water interfacial films." Environmental Science and
Technology 30: 1014-1022.
Prausnitz, J. M.; Lichtenthaler, R. N.; Azevedo, E. G. d. (1986). Molecular Thermodynamics
of Fluid-PhaseEquilibria.Englewood Cliffs, NJ, Prentice-Hall, Inc.
Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M. (1993). Environmental Organic
Chemistry. New York, NY, John Wiley & Sons, Inc.
Wauchope, R. D.; Getzen, F. W. (1972). "Temperature dependence of solubilities in water
and heats of fusion of solid aromatic hydrocarbons." Journal of Chemical Engineering
Data 17: 38-41.
227
I
li
Appendix B.
EVALUATION OF SOLID PHASE EXTRACTION METHODS FOR SEPARATING
DISSOLVED AND COLLOID-ASSOCIATED CONTAMINANTS IN GROUNDWATER
228
I
III~------Cy
~--r~---- r~--.^..^;-l- -^~i.rr---I--U-l~.,~~~-~~...x~,l~-~ .I ~,.. ---
---
Abstract
Reverse phase separation of dissolved and colloid-associated compounds was
investigated as a method to isolate in situ organic colloid-associated contaminants in
groundwater. Sep Pak cartridges (octadecyl bonded silica) removed > 98% of dissolved
polycyclic aromatic hydrocarbon (PAH) mass in a 300 mL volume of coal tar-equilibrated
water at a flowrate of 12 mL/min. At this flowrate, separation of humic acid-associated
PAHs was not quantitative. In addition to dissolved species, from 50% (chrysene) to 30%
(benzo(ghi)perylene) of colloid-associated mass was retained on the Sep Pak, suggesting that
colloid desorption occurred in the cartridge. Separation with Empore solid phase extraction
disks was also hindered by colloid desorption. Sep Pak reverse phase separation was applied
to groundwater from a coal tar site. Isolation of colloid-associated PAHs was unsuccessful
because of low organic colloid concentrations.
229
Introduction
Groundwater colloids can significantly enhance the transport of particle reactive
organic compounds over the transport of dissolved species. Two general methods have been
used to quantify the importance of colloid-mediated contaminant transport. First, researchers
have tried to isolate the colloid-phase and quantify the associated contaminant load with
separation techniques such as ultrafiltration, and field flow fractionation. These separation
methods suffer from many artifacts. For example, hydrophobic compounds are lost to organic
membranes and at low colloid concentrations, breakthrough to the dissolved phase occurs. A
quiescent colloid collection system has been developed whereby a sampling device collects
colloids which flow through the screened monitoring well. This method requires many
months for sample collection and it may be difficult to quantify organic contaminant
concentrations in the colloids. The second type of colloid characterization methods do not
require physical alteration of groundwater samples. Fluorescent probes are added to
groundwater samples to quantify colloid-water partition coefficients (Backhus and Gschwend,
1990; Gauthier et al., 1986). While this method demonstrates the potential for colloidfacilitated transport, it provides no measure of colloid-associated contaminant concentrations.
Fluorescent probes can also be dynamically quenched by the presence of dissolved species
(e.g., copper ions, oxygen (Lakowicz, 1983)) in the water sample, falsely indicating colloidassociation. There is a need for nonintrusive separation methods to isolate and quantify the in
situ colloid-associated contaminants in groundwater samples with a minimum of manipulation
artifacts.
Reverse phase separation is a method which allows separation of contaminants sorbed
to organic colloids with no sample preconcentration. A water sample flows through a solid
phase sorbent which sorbs the dissolved compound species. The colloid-associated
contaminants are carried through the sorbent to the effluent with the organic colloids. The
concentration of in situ colloid-bound contaminants present in the water sample can be
quantified by analysis of the effluent fraction.
The partition coefficients of model organic colloids (Landrum et al., 1984) and
colloids from natural waters (Backhus and Gschwend, 1990; Gauthier et al., 1986; Landrum et
230
-
I
I
al., 1984) have been quantified by reverse phase separation. In each of these cases, however,
a probe compound was added to the water sample before separation. The partition coefficient
was then quantified from the ratio of the compound mass in the effluent to that in retained by
the solid phase sorbent. Reverse phase separation has not been used to isolate in situ colloidassociated contaminants without the addition of a spike compound.
A reverse phase separation method for field applications must satisfy additional criteria
not necessary for these lab studies. Radio-labelled spiking compounds were used in the
laboratory studies (Gauthier et al., 1986; Landrum et al., 1984), and thus water volumes of
order 10 mL were sufficient to achieve separation. Organic compounds for which colloid
transport is important have low solubilities and so significantly larger volumes (100s mL) of
groundwater would be necessary to detect compounds in the effluent phase. The dissolved
compound mass in these large columns cannot exceed the solid phase sorbent capacity and
breakthrough to the effluent fraction. Sorbent pores may become clogged by particles and
prevent the passage of the organic colloid phase. Finally, the residence time in the column
must be short enough that little colloid-associated contaminant mass desorbs from the colloid
phase during groundwater contact with the sorbent. Partition coefficients determined by
reverse phase separation were shown to vary with organic colloid concentration (Landrum et
al., 1984) as a result of these mechanisms (i.e., colloid filtration, compound desorption).
These effects are likely to be exacerbated during the separation of large volume field samples.
While reverse phase separation may allow isolation of in situ colloid-associated contaminants,
it may only provide a lower bound of the concentration of colloid-associated contaminants in
groundwater samples.
In this study, several reverse phase separation systems were tested for their ability to
isolate colloid-associated polycyclic aromatic hydrocarbons (PAHs). Separation efficiencies
were evaluated with a humic acid solution containing a known amount of colloid-associated
compound mass. The reverse phase method was also applied to groundwater samples from a
coal tar site where transport of PAHs by organic colloids was strongly suspected (Chapters 2
and 3).
231
Methods
Chemicals
Humic acid (Aldrich, Milwaukee, WI) solutions were made up with sodium chloride
(Mallinckrodt, Paris, KY), sodium hydroxide and hydrochloric acid (Fisher Scientific,
Fairlawn, NJ). Organic carbon standards were made with potassium hydrogen phthalate
(Sigma, St. Louis, MO) and acidified with phosphoric acid (Mallinckrodt). Internal standards
of p-terphenyl, deuterated (d-) perylene, and d-benz(a)anthracene (Ultra Scientific, North
Kingstown, RI) were made in methanol (OmniSolve, EM Science, Gibbstown, NJ).
OmniSolve methylene chloride and hexane were also used.
Quantitative Breakthrough of Humic Acid
Quantitative passage of humic acid through the solid phase sorbents was first evaluated
with a humic acid solution containing no PAHs. The solution chemistry mimicked the
groundwater chemistry at Site YYZ, where reverse phase separation would be applied.
Aldrich humic acid (about 20 mg) was dissolved in 50 mL of 0.01 N sodium hydroxide. The
base solution was neutralized with the addition of an equal volume of 0.01 N hydrochloric
acid and diluted to 1.5 L. Sodium chloride salt (1.2 g) was added to bring the solution
conductivity to 1.05 mS, within the range of observed groundwater conductivities (Chapter 2).
The final solution pH was 7.2 and the organic carbon content was about 10 mgc/L,
representative of the levels of humic acid-like organic matter in the Site YYZ groundwater
(Chapter 3).
The solid phase sorbent systems were cleaned as described below. Purified water
(18 MQ, Aries purification system, Vaponics, Rockland, MA) was flushed through the reverse
phase separation system until the organic carbon content of the effluent did not differ from
the organic carbon of the 18 MQ water, within measurement error. Humic acid solution (50100 mL) was passed through the reverse phase system. The organic carbon content of the
starting solution and the effluent were measured to quantify the fraction of organic carbon
breakthrough for the reverse phase system.
Organic carbon was analyzed with a Shimadzu TOC-5000 Organic Carbon analyzer.
Aqueous samples were acidified with phosphoric acid and bubbled with argon or nitrogen for
232
..~.~~1.__~.,..
~_..-.-..IP...~I---
II*li^ _Il~fll.-~ ~-sllC-~
~. ~.-~1~--~I-_.__~
I_- 1~-.-I-~-~-C.-YL-~I~..
-~-~--~~~~~
5 minutes to remove inorganic carbon. The TOC analyzer was calibrated with external
standards of potassium hydrogen phthalate.
Reverse Phase Separations
The sorbent capacity and separation efficiency of the reverse phase systems were
evaluated with two model solutions containing polycyclic aromatic hydrocarbons. The first
solution was 18 Mn water containing 0.8 g/L sodium chloride. The second was a humic acid
solution as described above. Coal tar (3-4 mL) from the W40M monitoring well was added
to each of these solutions. The tar-containing solutions were stirred for 4 hours to aid
equilibration of PAHs between the tar and the aqueous solution. The solutions were
undisturbed for 5 days, allowing tar droplets to settle from the aqueous phase. Aliquots of the
tar-equilibrated solutions were then siphoned from the equilibration flasks. A piece of
aluminum tubing with a closed valve was primed with 18 MQ water and inserted in the
aqueous phase. The tube inlet was located well above the settled tar layer. The valve was
opened and four tube volumes of the equilibrated aqueous phase was allowed to flow to waste
before multiple 300 mL samples were taken in foil-wrapped BOD bottles.
One of the 300 mL samples for each tar-equilibrated solution was analyzed by liquidliquid extraction (LLE) to enable mass balance for the reverse phase separation procedure.
The aqueous sample was spiked with an internal standard of p-terphenyl and extracted three
times with 20 mL of methylene chloride. The extracts were combined and transferred to
hexane. The PAH fraction was isolated by silica gel chromatography according to the method
for field samples (Chapter 2).
The remaining 300 mL samples were separated with the reverse phase separation
systems. The aqueous solution was spiked with p-terphenyl and passed through cleaned solid
phase sorbents. The effluent was collected and its PAH content was quantified by the LLE
extraction procedure described above. The PAHs retained by the solid phase sorbent were
eluted with methylene chloride according to system-specific methods described below. The
solid phase extracts (SPE) were then transferred to hexane and underwent silica gel separation
of the PAH fraction according to the LLE method.
The PAH fractions for each of the solid-phase and liquid-liquid extracts were analyzed
by gas chromatography-mass spectrometry with an HP5890GC and an HP5972MS (Hewlett
233
I
Packard, Palo Alto, CA). Injections were made to a heated (280 0 C) split-splitless injector.
The split was opened 0.75 min after injection. The column was a 30 m HP-5 (Hewlett
Packard) capillary column with a film thickness of 0.25 gtm. Column flow was maintained at
1 mL/min by an electronic flow controller. The ionization voltage was 70 eV. The
temperature program began at 70 0 C with a 20 0 C/min ramp to 180 0 C and followed by a ramp
to 300 0 C at 6oC/min. The final temperature was held for 8 min. The detector was held at
280 0 C. Compounds were quantified with external standards and all peak areas were
quantified by selective ion monitoring.
Reverse Phase Separation Systems
Cartridges. Maxi-Clean (Alltech, Deerfield, IL) and Sep Pak Plus (Waters, Milford, MA)
solid phase extraction cartridges with octadecyl (C1 8) bonded silica were evaluated for their
efficiency to separate colloid-associated compounds from dissolved species. Aqueous
solutions were drawn through each of these cartridges in a vacuum filtration system. The
cartridges were inserted through a silicon stopper in the filter flask for collection of the
colloid-containing effluent. A 100 mL Luer-tipped syringe barrel served as the sample
reservoir. The filter flask was attached to a teflon vacuum pump with an aluminum line. The
vacuum was adjusted to maintain a flow rate of 12 mL/min through the cartridge.
The Maxi-Clean cartridges were only evaluated for humic acid breakthrough and did
not undergo any cleaning procedure other than flushing with 18 M.Q water to background
levels.
The Sep Pak cartridges were initially cleaned with three 10 mL aliquots of methylene
chloride, followed by 10 mL of methanol and 20 mL of 18 MQ water. After sample
collection, the Sep Pak was eluted with three 10 mL aliquots of methylene chloride.
Extraction disks. Empore solid phase extraction disks (3M, St. Paul, MN) were also
evaluated for use in the reverse phase separation method. These disks contained a C1 , sorbent
phase imbedded in a Teflon matrix. Although Empore disks were shown to retain pesticides
in a humic acid solution (Senseman et al., 1995), their low resistance to flow may allow
throughput of colloids and associated PAHs at high flowrates. The disks were placed on a
fritted glass filter holder and samples were drawn through with the vacuum filtration system
described above at a flowrate of 185 mL/min. The disks were cleaned by drawing several
234
-i----r-.~ I~.~.
-- c -~--~ ---^~I
~-rar*
~r~-r_----I.~. ....~.r
--I.IIXI----~--r*-----l--~P1-l ---
~P
~-----1-1-~I--~~-^-l
I~~-C--~ ~~ -~~--~-I~----~--~--
milliliters of methylene chloride through the disk and allowing it to soak for several minutes
before a total of 10 mL was pulled through the disk. The methylene chloride was followed
by a rinses of methanol (10 mL) and 18 Mn water (10 mL). After sample separation the
extraction disk was dried by drawing air through the disk for 5 min. The disk was then
eluted with three 10 mL aliquots of methylene chloride, allowing each to soak for several
minutes after passage of 1 mL.
Other Colloid Phases
The ability of the Sep Pak cartridges to pass colloid phases other than organic matter
was also investigated. A 1.5 mg/L kaolin (EM Science) solution with a turbidity of
1.2 + 0.04 NTU was passed through the Sep Pak in 50 mL increments. The effluent turbidity
was measured after each increment until a total of 500 mL had been passed. Turbidity
measures were made with an HF Scientific (Ft. Myers, FL) turbidimeter. Tar retention by the
Sep Pak was also tested by adding a 20 gLL droplet to 2 mL of water in the reservoir. This
water was pulled through the cartridge, followed by an additional 50 mL.
Field Application
The Sep Pak reverse phase separation method was applied to groundwater samples
from Site YYZ to quantify colloid associated contaminants. Triplicate groundwater samples
were collected in foil-wrapped 300 mL BOD bottles by slow pumping methods (Chapter 2).
One bottle was retained for analysis of the total dissolved plus colloid-associated polycyclic
aromatic hydrocarbons by LLE at MIT. The second sample was spiked with an internal
standard containing p-terphenyl and d12-perylene and separated in the field with a pre-cleaned
Sep Pak system. Photodegradation of PAHs was minimized by foil-wrapping the system to
the greatest extent possible. After separation, the effluent was transferred to a clean BOD
bottle and liquid-liquid extracted at MIT after spiking with a recovery standard of
d-benz(a)anthracene. The third groundwater sample was returned to the laboratory for Sep
Pak separation in the event that the contamination was introduced during field manipulations.
Contamination introduced in the field was quantified by reverse phase separation of a 300 mL
aliquot of 18 M92 water brought from MIT in a BOD bottle.
The vacuum filtration system was cleaned in the laboratory before use in the field. A
new filter flask was used for effluent collection at each well. The filter flasks were soaked in
235
I
-r-
chromic-sulfuric acid solution overnight, followed by rinsing with 18 Mn2 water, methanol
and methylene chloride. The individual Sep Paks were pre-cleaned in the laboratory and
wrapped in foil. The foil-wrapped cartridges were then wrapped with moist paper and placed
in a Zip Loc bag to maintain the moisture content in the cartridge until use 48 h later. After
sample separation in the field, this same storage procedure was used to return the cartridges to
the laboratory. Twenty four hours after collection, the Sep Paks were hand flushed with
4 mL of methanol in a syringe to remove residual water. The Sep Paks were then eluted by
gravitational flow of 50 mL of methylene chloride contained in a glass syringe with no
headspace.
All methylene chloride extracts from the field samples were transferred to hexane,
chromatographed on silica gel, and analyzed by GC/MS.
236
-._~~~~_.e~,~,~.~_~, -i-rrr-rarrrrcll---~--p-~- r^r~ ---~ritCII
~
--4CP~--~1^1~~ ........._-~- I- 1-~~~
I
Results and Discussion
Evaluation of Humic Acid Passage by Reverse Phase Separation Systems
The solid phase extraction systems were first evaluated for their retention of organic
humic acid colloids. Only those systems which quantitatively passed humic acid, as measured
by the ratios of organic carbon concentration in the effluent to that in the starting solution,
were evaluated for reverse phase separation of colloid-associated polycyclic aromatic
hydrocarbons. Maxi-Clean cartridges were eliminated from subsequent reverse phase
evaluations due to their retention of humic acid (Table B.1). In addition to the low organic
carbon concentration, the effluent lacked coloration in comparison to the starting humic acid
solution. The retained humic acid was observed as a brown band at the head of the cartridge.
The results of the humic acid evaluation showed the Empore extraction disks and the
Sep Pak cartridges to be promising candidates for reverse phase separation of colloidassociated PAHs. High organic carbon breakthrough was observed for the Empore disks, but
some disk discoloration was observed, indicating some retention of humic acid. The Sep Pak
cartridges had complete passage of organic carbon and showed no discoloration from humic
acid retention in the porous sorbent.
The Sep Pak cartridges were also evaluated for their retention of clay particles and tar.
After passage of only 50 mL of a 1.5 mg/L kaolin solution (1.2 NTU), the effluent turbidity
was 0.2 NTU. The effluent turbidity remained constant at 0.1 NTU for the next 500 mL
collected. Therefore, over 90% of clay particles were filtered by the porous sorbent media.
Tar was also retained by the Sep Pak, as expected from the hydrophobic nature of the packing
material. A 20 CpL tar droplet was not forced through the column by additional passage of
water. The effect that pore clogging by clay particles or tar droplets might have on the
passage of organic carbon macromolecules through the sorbent cartridge was not investigated.
The separated fraction containing the colloid-associated contaminants can be postulated
for various colloid types. If the contaminants were sorbed to suspended organic colloids, the
contaminant mass retained on the Sep Pak would be less than extracted from an
unfractionated water sample of equivalent volume. The difference in contaminant mass
between that eluted from the Sep Pak and that contained in the extract of unfractionated water
237
___~_______-----
---
Table B.1. Humic acid passage through solid phase extraction systems.
Solid phase
extraction system
Initial humic acid
concentration
Effluent humic acid
concentration
Humic acid
breakthrough
(mg C/L)
(mg C/L)
(%)
Empore disk
8.7
9
8.3
7.3
95
80
Maxi-Clean
8.7
4
45
12.5
12.8
100
cartridge
Sep Pak cartridge
238
I
-
L
would be found in the effluent extract. If the contaminants were associated with clay particle
colloids or tar droplets, they would be retained on the Sep Pak, in addition to the contaminant
mass dissolved in the water sample. In this case the mass of compounds eluted from the Sep
Pak would by equal to that in an extraction of an equal volume of unfractionated water. No
contaminant mass would be detected in the effluent.
Evaluation of Reverse Phase Separation of Colloid-Associated PAHs
Two model solutions were used to evaluate the separation of dissolved and colloidassociated polycyclic aromatic hydrocarbons by the reverse phase separation systems. Tar
equilibrated water, containing only dissolved species, was used to evaluate the extraction
capacity of the sorbent. While the mass of solid phase sorbent (300 mg) greatly exceeded the
total dissolved compound mass (5 mg) in the volumes tested, the flowrates employed may not
have allowed sufficient time for uptake of dissolved compounds by the sorbent. Thus, the
effluent would also contain dissolved compound mass, an artifact of too fast flow rates. Once
certain that dissolved compounds would be sorbed by the solid phase sorbent, a tarequilibrated humic acid solution was used to quantify the efficiency of separation of the
colloid-associated mass from the dissolved mass. The total colloid-associated mass was
known by comparison with tar-equilibrated water. Thus, the separation efficiency could be
evaluated by comparison of the mass of compounds in the effluent with the known colloidassociated mass in the starting solution.
The concentrations of polycyclic aromatic hydrocarbons in the humic acid solution
were enhanced over those in tar-equilibrated water (Table B.2). The solubility enhancement
by the humic acid increased with compound hydrophobicity, from no enhancement of pyrene
and fluoranthene to an enhancement of 3 for benzo(ghi)perylene. For each compound, the
fraction of dissolved mass in the humic acid solution was calculated by dividing the
compound concentration in tar-equilibrated water with that in the humic acid solution. This
fraction is the portion of total compound mass expected to be eluted from the solid sorbent
after reverse phase separation if the separation of the colloid-associated mass is ideal. The
remainder of the compound mass would be present with the organic colloids in the effluent.
239
---C
Table B.2. Polycyclic aromatic hydrocarbon concentrations in tar-equilibrated water and a
tar-equilibrated 7 mgc/L Aldrich humic acid solution. Concentrations were quantified by
methylene chloride extractions of unfractionated samples.
Compound
18 Mn water
(mg/L)
Humic acid
solution
(mg/L)
Ratio:
Humic sol'n
18 Mn water
Ratio:
Dissolved
Diss'd+Colloid
0.012
0.013
1.1
0.9
Pyrene
(PY)
0.012
0.0079
0.7
0.4
Benz(a)anthracene
0.0021
0.0048
2.3
0.4
0.0017
0.0038
2.2
0.4
0.00057
0.0016
2.8
0.4
Benzo(a)pyrene
(BaP)
0.001
0.0029
2.9
0.3
Indeno(123cd)pyrene
0.00035
0.001
2.9
0.3
0.00029
0.00092
3.2
0.3
Fluoranthene
(FL)
(BA)
Chrysene
(CH)
Benzo(e)-
pyrene
(BeP)
(IP)
Benzo(ghi)perylene
(BP)
240
I_11I1_--~-~I I~I-q-.-.~-^l -I~II_
~--..^~XI1P------C
--.
III-LI-l(rlllYI*^__^LI
1_
Greater than 98% of the dissolved compound mass was sorbed by the Sep Pak
cartridges at the flowrate of 12 mL/min. A mass balance was performed for the reverse phase
separation by summing the mass of each compound in the solid phase extract (SPE) and the
effluent. Sixty to eighty percent of compound mass was recovered in the reverse phase
separated fractions in comparison to the liquid-liquid extraction of a duplicate volume of
water (Table B.3). The recoveries of benzo(a)pyrene and benzo(ghi)perylene may have been
improved with the addition of deuterated perylene internal standard. The fraction of
compound mass in the effluent fraction was only 1-2% of the total mass of PAHs in the tarequilibrated water. Additionally, less than 0.5% of the p-terphenyl internal standard added
before reverse phase separation was extracted from the effluent fraction. Therefore, the
flowrate for reverse phase separation allowed sufficient contact time for the dissolved
compound mass to be exchanged from the aqueous solution to the solid sorbent.
A portion of the colloid-associated PAHs in the tar-equilibrated humic acid solution
was separated from the dissolved compound mass by reverse phase separation with the Sep
Pak cartridges. The mass balance for this system was also close to 100% by comparison with
a liquid-liquid extraction of a duplicate sample volume of the humic acid solution (Table B.4).
Detectable levels of PAHs were found in the effluent and the compound masses eluted from
the solid phase sorbent were greater than the dissolved mass in the humic acid solution.
These results demonstrate that colloid-associated compounds were carried through the Sep Pak
with the colloid phase. For each compound, the ratio of mass eluted from the solid sorbent
(SPE) to the total mass in the humic acid solution (Table B.4) was greater than the calculated
fraction of dissolved compound mass in the starting solution (Table B.2). Thus, PAHs that
were colloid-associated in the starting solution were retained by the solid sorbent, in addition
to the initially dissolved species. The masses of compounds in the Sep Pak in excess of
dissolved levels may have been retained with colloids filtered from solution by the porous
media or may have desorbed from the colloids in response to the gradient induced by solid
phase sorbent uptake of dissolved compounds.
241
~
Table B.3. Sep Pak separation of dissolved and colloid-associated polycyclic aromatic
hydrocarbons in tar-equilibrated 18 MC water.
Compound
SPE
Effluent
(mg/L)
(mg/L)
FL
0.0096
0.000 025
0.8
0.8
0.002
PY
0.0064
0.000 019
0.5
0.5
0.003
BA
0.0015
0.000 014
0.7
0.7
0.01
CH
0.0013
0.000 012
0.8
0.8
0.01
BeP
0.0004
BaP
0.0007
IP
0.0002
BP
0.000 17
Mass
Balancet
Ratio:
SPE
Breakthrough
18 M9 H 2 0
0.7
0.000 01
0.8
0.7
0.02
0.6
< 0.000 006
0.7
t (SPE + Effluent)/18 MR water
Effluent/(SPE + Effluent)
242
0.6
0.03
- .xI.-.~,
1
-~"~-~ ~311~1- --. ,
..,~ .~.~,~~. .~..~~ I~-~----~---c~--------
-- I
Table B.4. Sep Pak separation of dissolved and colloid-associated polycyclic aromatic
hydrocarbons in tar-equilibrated humic acid solution.
Mass
Balancet
Ratio:
SPE
Humic
acid
solution
Breakthrough*
Fraction
colloidassociated
mass
retained
SPE
Effluent
(mg/L)
(mg/L)
FL
0.019
0.002
1.6
1.5
0.1
PY
0.013
0.0016
1.8
1.6
0.1
BA
0.0037
0.0016
1.1
0.8
0.3
0.5
CH
0.0029
0.0013
1.1
0.8
0.3
0.5
BeP
0.000 92
0.000 78
1.1
0.6
0.5
0.2
BaP
0.0017
0.0013
1
0.6
0.4
0.4
IP
0.000 43
0.000 43
0.9
0.4
0.5
0.3
BP
0.000 43
0.000 42
0.9
0.4
0.5
0.3
Compound
t (SPE + Effluent)/Humic acid solution
Effluent/(SPE + Effluent)
243
~
The efficiency of reverse phase separation was quantified by calculating the fraction of
initially colloid-associated compound mass retained by the Sep Pak (Table B.4). For
example, 60% of the total chrysene mass was colloid-associated in the humic acid solution,
but only 30% of the total mass was detected in the effluent. Thus, 50% of the initially
colloid-associated chrysene was retained by the Sep Pak, in addition to the dissolved
chrysene. For PAHs with increasing hydrophobicities, decreasing fractions of the initially
colloid-bound masses were retained by the solid sorbent. If colloids with their associated
compound mass were strained from solution, a constant retention would have been observed
for all PAHs regardless of hydrophobicity. Colloid straining was likely not an important
separation artifact because the fractions of colloid-associated mass retained by the Sep Pak
were not constant for all PAHs. If cartridge residence times were long enough for colloidassociated compounds to desorb from the colloid phase, greater fractions of compounds with
low hydrophobicities would be retained than for compounds with high hydrophobicities.
Uptake of dissolved compounds by the solid phase sorbent induces a gradient for desorption
from the colloid phase to maintain phase equilibrium. Low hydrophobicity compounds must
desorb more mass than do very hydrophobic compounds to maintain this equilibrium. Thus,
smaller fractions of the initially colloid-associated mass with increasing compound
hydrophobicity would be retained by the solid phase sorbent, as observed for reverse phase
separation of the humic acid solution. Therefore, reverse phase separation with Sep Pak
cartridges does allow qualitative identification of organic colloid-associated compounds, but
quantification of their concentrations is limited by cartridge residence times greater than
colloid desorption times.
Reverse phase separation of the humic acid solution with the Empore extraction disks
was also limited by desorption of colloid-associated mass. The ratios of SPE masses to total
masses in the humic acid solution (Table B.5) were greater than the fraction of dissolved
compound mass calculated in the humic acid solution (Table B.2). The fraction of initially
colloid-associated mass retained by the extraction disk was not constant for all PAH, but
decreased for compounds with increasing hydrophobicity, suggesting colloid desorption.
244
I---
...,. _-..-^
I----^1
~
l~rrrr~*--r~..----^---llc-Il--yr~--l~,.
- --.-Y--
-rv--- ~~-I
~sl'^-~-rr----
r
I
Table B.5. Empore disk separation of dissolved and colloid-associated polycyclic aromatic
hydrocarbons in tar-equilibrated humic acid solution.
Fraction colloidassociated mass
retained by disk
Compound
Solid phase
extraction
(mg/L)
Ratio:
SPE
Humic acid sol'n
PY
0.011
1.4
BA
0.004
0.8
0.7
CH
0.0026
0.7
0.5
BaP
0.0017
0.6
0.4
IP
0.0056
0.6
0.4
BP
0.0043
0.5
0.3
245
---
Although the flowrate was fifteen times higher than through the Sep Paks, the disk in the
filter holder remained exposed to the sample reservoir throughout the course of the separation.
This also allowed for compound desorption from the colloid phase local to the disk, elevating
the masses of compounds sorbed by the sorbent disk above dissolved levels. For all PAHs,
the fraction of initially colloid-associated mass retained in this reverse phase separation
system was similar to the fractions retained by the Sep Paks.
There was little difference in separation efficiency of colloid-associated PAHs between
the Sep Pak cartridge and Empore disk systems, but sample separation was easier with the
cartridge system. The extraction disks were difficult to clean and elute in the vacuum
filtration system without drying the disks. The Sep Pak sorbent was contained in a cartridge
which minimized sorbent contamination by handling and the cartridges also had Luer fittings
which enabled elution with a syringe containing no head space. For these reasons the Sep
Pak cartridge system was chosen to investigate in situ colloid-associated PAHs in
groundwater.
Reverse Phase Separation of Colloid-Associated PAHs in Groundwater
The reverse phase separation technique was applied to groundwater samples from a
coal tar site in order to isolate the colloid-associated polycyclic aromatic hydrocarbons.
Previous investigations at this site had yielded groundwater PAH concentrations greater than
dissolved levels in equilibrium with coal tar (Chapter 2). This concentration enhancement
suggested the presence of a colloid phase facilitating PAH transport in groundwater at this site
and fluorescence quenching studies supported the hypothesis that humic acid-like organic
colloids enhanced groundwater PAH solubilities (Chapter 3). Groundwater samples were
collected from wells (W20S, W100S, W100M) which had exhibited enhanced PAH
concentrations at some time. Monitoring well W40M which always had elevated PAH
concentrations could not be investigated due to a large quantity of tar in the well at the time
of sampling. Representative groundwater samples could not be taken from this well without
the inclusion of tar artifacts.
Polycyclic aromatic hydrocarbon concentrations were not elevated above tarequilibrium levels at any of the sampled wells (Figure B.1). Thus, reverse phase separations
246
I
~ac ---------
1
~111111~1
W20S
le-1
*
le-2 o
LLE
Sep Pak
le-3
0
le-4
le-5
le-5
le-4
le-3
le-2
le-1
Measured Equilibrium (mg/L)
W100S
W100M
le-1
le-1
le-2
le-2
le-3
le-3
le-4
le-4
' " " ' """ ' - ' ", - -""'
1e-5
le-1
le-4
le-3
le-2
le-5
le-5
V
le-5
le-4
le-3
le-2
le-1
Measured Equilibrium (mg/L)
Measured Equilibrium (mg/L)
Figure B.1. Polycyclic aromatic hydrocarbon concentrations in groundwater in June 1997.
Concentrations determined by liquid-liquid extraction and the Sep Pak extracted
concentrations from reverse phase separation are reported. Concentrations that were less than
detection limits are indicated with an arrow and a symbol at the detection limit.
247
--
of groundwater samples were not expected to contain detectable levels of colloid-associated
PAHs in the effluent fraction. The effluent compound mass accounted for less than 0.5% of
the total compound mass at all wells (Table B.6a,b,c). From the detection limits of PAHs in
the effluent fraction, an upper bound of 1 mgc/L of organic colloids was estimated to be
present to enhance PAH concentrations in the groundwater. For example, if the humic acid
colloids were representative of the groundwater organic colloids, only 60% of the colloidassociated benzo(a)pyrene in the initial groundwater sample would be present in the effluent.
Assuming groundwater organic colloids had partition coefficients similar to Aldrich humic
acid (as suggested by the fluorescence quenching studies (Chapter 3)), the mass of
benzo(a)pyrene associated with organic colloids present at levels less than 1 mgc/L would
have been below the detection limits for the effluent analysis. The amount of organic carbon
in the effluent which was capable of quenching pyrene fluorescence, and hence enhancing
PAH concentrations in equilibrium with tar, was not quantified because the Sep Pak had not
been rinsed with 18 MO water prior to use, nor was the amount of fluorescence-quenching
material leached from the Sep Pak known.
248
-ur*-I-~-----LI
.Il--...~---..~LI1~---I-^^I
1II1I---.~--1
CI-~-YYL
_IIIII-^
^Ill
1)1
_.~-~_lsl__^_
~-r~l
~_i~_
Illllll~
- I^C-
I
_~
Table B.6a. Sep Pak separation of dissolved and colloid-associated polycyclic aromatic
hydrocarbons at monitoring well W20S.
Compound
SPE
(mg/L)
Effluent
(mg/L)
Breakthrough
FL
0.0018
0.000 04
< 0.001
PY
0.000 54
0.000 02
< 0.001
BA
0.000 11
< 0.000 03
< 0.001
CH
0.000 12
< 0.000 03
<0.002
BaP
< 0.000 07
< 0.0001
BP
< 0.000 04
< 0.000 05
Table B.6b. Sep Pak separation of dissolved and colloid-associated polycyclic aromatic
hydrocarbons at monitoring well W100S.
Compound
SPE
(mg/L)
Effluent
(mg/L)
Breakthrough
FL
0.0098
0.000 025
0.001
PY
0.029
0.000 032
0.001
BA
0.000 33
0.000 029
0.002
CH
0.000 27
0.000 028
0.004
BaP
< 0.0003
< 0.000 03
BP
< 0.000 01
< 0.000 002
249
-ir
Table B.6c. Sep Pak separation of dissolved and colloid-associated polycyclic aromatic
hydrocarbons at monitoring well W100M.
Compound
SPE
(mg/L)
Effluent
(mg/L)
Breakthrough
FL
0.0038
0.000 026
< 0.001
PY
0.0038
0.000 028
< 0.001
BA
0.000 56
0.000 024
< 0.001
CH
0.000 56
0.000 023
0.001
BaP
0.000 36
< 0.000 02
< 0.05
BP
0.000 14
< 0.000 001
< 0.08
250
- I
IL
111-1--71111~
11 I
-I
I
Conclusions
The method development indicates that reverse phase separation may be a promising
way to isolate organic colloid-associated compounds from field samples with a minimum of
sample manipulation. Presently this method only allows verification of the presence of
organic-colloid associated compounds by their detection in the effluent fraction.
Quantification of the fraction of colloid-associated mass, and hence groundwater-colloid
partition coefficients, is hindered by colloid desorption while the sample is in contact with the
solid phase sorbent. Additionally, compound detection limits may restrict the application of
reverse phase separation to field sites with high mobile organic colloid concentrations.
Further investigation is required to quantify the effects of other colloid phases, such as clays,
on reverse phase separation of organic colloids and their associated contaminant loads.
251
References
Backhus, D. A.; Gschwend, P. M. (1990). "Fluorescent polycyclic aromatic hydrocarbons as
probes for studying the impact of colloids on pollutant transport in groundwater."
Environmental Science and Technology 24: 1214-1223.
Gauthier, T. D.; Shame, E. C.; Guerin, W. F.; Seitz, W. R.; Grant, C. L. (1986).
"Fluorescence quenching method for determining equilibrium constants for polycyclic
aromatic hydrocarbons binding to dissolved humic materials." Environmental Science
and Technology 20: 1162-1166.
Lakowicz, J. R. (1983). Principles of Fluorescence Spectroscopy. New York, Plennum Press.
Landrum, P. F.; Nihart, S. R.; Gardner, W. S.(1984). "Reverse-phase separation method for
determining pollutant binding to Aldrich humic acid and dissolved organic carbon of
natural waters." Environmental Science and Technology 18: 187-192.
Senseman, S. A.; Lavy, T. L.; Mattice, J. D.; Gbur, E. E. (1995). "Influence of dissolved
humic acid and Ca-montmorillinite clay on pesticide extraction efficiency from water
using solid-phase extraction disks." Environmental Science and Technology 29:
2647-2653.
252
--
I
III~L~PIY~ C
I
,,
Appendix C.
CALCULATION OF THE EFFECTIVE DIFFUSION COEFFICIENT IN
DOUGLAS FIR
253
This appendix details the calculation of the effective diffusion coefficient in Douglas
fir (Burr and Stamm, 1947, Figure6.4, Chapter 6). Experimental results of the monoaromatic
hydrocarbon uptake in wood suggested that only tangential diffusion was important to overall
compound uptake (Chapter 6). Therefore, only calculations of the tangential diffusion
coefficient are shown here. The methodology is the same for calculating longitudinal wood
diffusion, however, the individual parameters in the resistance terms have different values due
to the differing orientation of the wood fibres relative to the diffusion pathway.
The diffusion coefficient of a compound in wood (Deff) is related to its aqueous
diffusivity (D,) by the overall resistance of the physical wood structure:
Deff
R 'JDw
=
where RToT is the overall resistance of the physical wood structure given as the sum of
individual resistances (refer Figure 6.4, Chapter 6):
1
RTOT
-
1
Rf
1
+
R
+
nT
1
a
1
R
e
(2)
1
Rb +
1
d + Rc
where R, is the fibre cavity resistance in the fibre direction;
Rb
is the pit cavity resistance;
Rc is the resistance of the pit membrane substance;
Rd is the pit membrane pore resistance;
R, is the resistance of the overlapping cell walls;
Rf is the resistance of the continuous cell wall, and
nT is the number of cells per unit length in the fibre direction.
254
~x~_*-~-~l-lr.--~
-~e~^~n...l.
l i .-
L111~"4
~"IIX-----~------_l____l___
I_-II^-~-^---~I
Il-ll-L_^
_ _.^
~yC-(-l
I
These equations can also be written in terms of conductance (= resistance'):
DwoD = CTOTDw
(3)
CTOT = Cf +
1
(4)
Ce +
1
e
cb
l
1
Cd +c
where the conductances, C,, are for the physical structures as described for Eq. 2.
The following expressions are used to calculate the individual conductances (Behr et al.,
1953):
Ca= 1
qP
Cb-
(5)
(6)
LM
qm
(7)
CC-
qT
C
_-
QT
L
Cf = Lm nTST
where the parameters are defined in the following table.
255
(8)
(9)
(10)
Parameter
Description
Value for
Douglas fir
nT
average number of fibres traversed per cm in the
300
tangential direction
L
average thickness of the double cell walls of
0.000 92
swollen wood
L
average pit membrane thickness
0.0001
qp
fractional area of tangential fibre walls covered by
0.014
pits
fractional cross-sectional area of the transient pit
qm
membrane-capillaries of solvent-swollen wood for
0.000 11
the passage of molecules the size of water
molecules, exclusive of permanent pores
effective fractional cross-sectional area of
qT
permanent pit-membrane pores for transverse
0.000 52
passage
fractional cross-section of the transient cell wall
capillaries of solvent-swollen wood effective for
QT
the passage of molecules the size of water
molecules in the transverse direction from one
0.0063
cavity to another
fractional cross-section of the transient cell wall
S,
capillaries of solvent-swollen wood effective for
passage of molecules the size of water molecules
in the tangential direction
256
0.0078
__
_~_____~__~_~~~;_;_~_~~__ _I~~.,.Y.^~----r~-.-r;.i ~i~-ur..ra---~-I-^_rrn-^r-----l
--11I~
~I
b
--
-
------------
These numerical values were used to calculate the individual and overall conductances for
Douglas fir shown in Figure 6.11 a (Chapter 6). The effect of equilibrium microscale
partitioning is hypothesized to increase the effective pathlength through any wood tissue
structure. The individual resistances were recalculated in Figure 6.11b with tissue partitioning
with lengths Lm and Lp multiplied by a retardation factor of 10.
References
Behr, E. A.; Briggs, D. R.; Kaufert, F. H. (1953). "Diffusion of dissolved materials through
wood." Journal of Physical Chemistry 57: 476-480.
Burr, H. K.; Stamm, A. J. (1947). "Diffusion in wood." Journal of Physical and Colliod
Chemistry 51: 240-261.
257
Appendix D.
ADDITIONAL TIME COURSE PLOTS OF MONOAROMATIC COMPOUND
UPTAKE BY WOOD
258
-7---~~11 P--
-~~--
850
800
0
750
O
A
*
replicate 1
replicate 3
replicate 2
control
700
650
600
550
0
-0
1500
1000
500
0
2000
2500
3000
3500
10000
000
Time (min)
800
750
O
O
A
*
0
o
0.00.
A
550 L
500
450
400
replicate 1
Sreplicate 3
A replicate2
*
control
o
0O
A
0
A
AO
0
O
0
A
350
1000
Time (min)
Figure D.1. Benzene uptake by Ponderosa pine chips.
259
10000
I^~I--~~~--~- -~-^~
zuu~
00
150
000
100
*
Control
Chips
o
0
500
1500
1000
2000
2500
3000
Time (min)
0
0
0
0
o
Control
Chips
S.
.
.
.
.
.I
.
.
.
.
.
.
.
100
Time (min)
Figure D.2. O-xylene uptake by Ponderosa pine chips.
260
.
.
.
..
I
1000
1
IC
-.-----I~l~--~--
~ _____~ _sll-l~--^L-
ID~--II^.
.- l .lll-~-L- ...- -~-
600 550
500
450
400
350
0
1000
500
1500
2000
2.500
3000
3500
4000
Time (min)
850
800
750
700
650
600
550
500
0[
0
sticks
control
450
400
'
350
1
10
'
100
Time (min)
Figure D.3. Benzene uptake by Douglas fir sticks.
261
1000
10000
*
A
500
Control
Sticks
4504
400,
-0
350
-A
300
-A
250
200
1
150
100
0
200
300
"y/~L
....
Illillllllil
5000
10000
15000
Time (min)
S
*S
250
-
*
Control
A
Sticks
0
*
A
200
A
iA
1000
Time (min)
Figure D.4. Toluene uptake by Douglas fir sticks.
262
10000
I-i~r.
~, __^~,--..~_r-I--r~
~l ----~---
------- -Ill~"'~r~--~
-TT -~'~-~III--~----~r~
-~~
~
7/
180
160
0
120
o
100
*
o
Control
Sticks
I..
iiiii
.. l,,I ,,l,,
Il,
.......................
-
2000
1000
3000
4000
5000
6000
7000
800023000
Time (min)
180
160
0
0
40
0
0 00
0
0
120
100
80
*
o
,
Control
Sticks
r
ii, i
o
0
i i, ,
,ll,
,
, ,r,11
1
1000
Time (min)
Figure D.5. O-xylene uptake by Douglas fir sticks.
263
,,00
10000
'-'ls~-~LI
1IP-l~------- -~-*_
264
~
~--
Appendix E.
REPRESENTATIVE GAS CHROMATOGRAMS
265
Figure E.1. W40M coal tar (1/1250 dilution).
Note the attenuation changes during the temperature program.
The pyrene peak corresponds to a tar concentration of 9300 mg/L. Other concentration
values are given in Table 2.2 (Chapter 2).
Peaks are identified with the following compound abbreviations:
NA - naphthalene
MeNA - methylnaphthalenes
C,NA - ethyl and dimethylnaphthalenes
PH - phenanthrene
AN - anthracene
MePH - methylphenanthrenes and methylanthracenes
PY - pyrene
FL - fluoranthene
BA - benz(a)anthracene
CH - chrysene
BaP - benzo(a)pyrene
BeP - benzo(e)pyrene
BP - benzo(ghi)perylene
IP - indeno(123-cd)pyrene
266
oG(,1
-4
1281~ P).I
au
?^6 C
(£t
30cl)
>3
TemP
Figure E.2. Oil isolated from the B4 boring.
268
269
vz
IkA*~
JAI
19~
Figure E.3. Pentane/acetone extract from the B4 core. Note the prevalence of polycyclic
aromatic hydrocarbon peaks, reflecting the nonaqueous phase liquid coal tar in these solids.
Compound abbreviations are given in Figure E.1.
270
IMi- A4J .JA
loNVo
9fii-
c
4,
1wr
ObT
271